Introduction

Bread wheat is one of the most important staple crops worldwide. Genetic improvement of wheat yield potential has been made in many countries in Europe (Austin et al. 1980; Brancourt-Hulmel et al. 2003), North America (McCaig and DePauw 1995; Donmez et al. 2001), South America (Ortiz-Monasterio et al. 1997) and Oceania (Siddique et al. 1989) over the past few decades. Significant genetic progress has also been made in improving yield potential of Chinese wheat cultivars during the past 50 years (Zhou et al. 2007; Zheng et al. 2011; Xiao et al. 2012; Gao et al. 2017). However, the gradual decrease in farmland, rapid increase in global population, decline in available water and climate change are important factors necessitating yield improvement. Thus, increased grain yield (GY) is the main goal in wheat breeding. Considering the complex, polygenic inheritance, low heritability, and significant influence of environment, yield improvement continues to be a huge challenge.

Wheat GY comprises three main components, viz. spike number per unit area (SN), kernel number per spike (KNS) and thousand-kernel weight (TKW). Increases in grain yield were mainly associated with increased components as well as harvest index (HI) (Donmez et al. 2001; Brancourt-Hulmel et al. 2003). Kernel, spike, plant height (PH) and flag leaf-related traits can also affect GY through impacting yield components. Among the three yield components, SN and KNS are more easily influenced by environment. Numerous quantitative trait loci (QTL) for these two traits have been identified, and one gene TaTEF-7A related to KNS was cloned (Cuthbert et al. 2008; Deng et al. 2011; Lee et al. 2014; Zheng et al. 2014; Gao et al. 2015; Sun et al. 2017). Previous studies showed higher heritabilities of TKW than SN and KNS, varying from 0.59 to 0.80 (Xiao and He 2003). TKW is determined by kernel dimensions, such as kernel length (KL) and kernel width (KW) (Zhang et al. 2014a; Huang et al. 2015). More than 100 QTL for TKW have been mapped on all 21 wheat chromosomes (Gegas et al. 2010; Ramya et al. 2010; Gao et al. 2015; Sukumaran et al. 2015; Wu et al. 2015; Sun et al. 2017), and over 20 kernel weight-related genes were cloned (Ma et al. 2010; Jiang et al. 2011; Su et al. 2011; Yang et al. 2012; Zhang et al. 2012b; Chang et al. 2013; Guo et al. 2013; Chang et al. 2014; Dong et al. 2014; Qin et al. 2014; Zhang et al. 2014b; Hanif et al. 2015; Wang et al. 2015; Ma et al. 2016; Wang et al. 2016b; Zhang et al. 2017c). KNS and TKW can be defined by spike length (SL) and dry weight (SDW) that are easier to select in small plots in wheat breeding. Genetic studies on these two traits were not given enough attention previously, particularly SDW. Heading date (HD), which is critical for optimal crop adaptation, yield potential and stability, is controlled by genes for vernalization (Vrn), photoperiod (Ppd) and earliness per se (Eps). Previous studies indicated that HD associated genes were distributed on almost all 21 chromosomes and expressed at different developmental stages (Griffiths et al. 2009; Huang et al. 2011; Zhang et al. 2012a; Takenaka and Kawahara 2013; Zikhali et al. 2014; Muterko et al. 2016).

Following deployment of the Green Revolution genes (Rht-B1b and Rht-D1b) much success was achieved in breeding wheat varieties with increased lodging resistance, higher HI, and higher yield potential (Peng et al. 1999; Ellis et al. 2005). To date, 24 Rht genes have been identified; Rht-B1b, Rht-D1b and Rht8 are the three dwarfing genes that are widely used in wheat cultivars (Peng et al. 1999; Hedden 2003; Ellis et al. 2005; Chen et al. 2015; Tian et al. 2017). Recently, Rht24 was identified on chromosome 6AL. This gene was widely distributed in elite wheat varieties, and reduced PH by 6.0–7.9 cm (Tian et al. 2017; Würschum et al. 2017). Varieties with long uppermost internode length (UIL) are preferred in wheat breeding, and related genes were mapped (Yu et al. 2014; Zhang et al. 2017b). Previous studies and production practice indicated that PH had a significant influence on yield-related traits (Peng et al. 1999; McCartney et al. 2005; Cuthbert et al. 2008; Gao et al. 2015).

Flag leaf area, including flag leaf length (FLL) and width (FLW), does not only affect photosynthetic performance (Xu and Zhao 1995; Sharma et al. 2003), but also determine population structure and impact plant stress response (Sourdille et al. 2002; Perez–Perez et al. 2010). Thus, flag leaf area is significantly associated with GY. The TaFLW1 gene associated with FLW was finely mapped close to Fhb5 on chromosome 5A (Xue et al. 2013). Although several QTL for FLL and FLW have been identified, it is necessary to identify more common and stable QTL for marker-assisted selection (MAS) in wheat breeding (Jia et al. 2013; Xue et al. 2013; Wu et al. 2016).

The International Wheat Genome Sequencing Consortium (IWGSC 2014) identified more than 124,000 genes in bread wheat. However, only about 65 important genes have been cloned and 150 functional markers were developed for these genes (Liu et al. 2012; Rasheed et al. 2016; Nadolska-Orczyk et al. 2017). Although numerous yield-related QTL have been identified (Azadi et al. 2015; Cui et al. 2014; Gao et al. 2015; Sun et al. 2017), there is still a large gap between the discovery of QTL, gene cloning and MAS due to the low heritability of related traits, limited number of genetic markers and the narrow genetic background of single mapping populations in previous studies (Nadolska-Orczyk et al. 2017).

Single nucleotide polymorphisms (SNP) are abundant in plant and animal genomes (Wang et al. 2014). SNPs are now widely used in QTL mapping and genome-wide association studies (GWAS) for economically important crops and animals (Wiedmann et al. 2008; Ganal et al. 2011; Song et al. 2013). The rice (44 K) and maize (50 and 55 K) SNP arrays are widely used in detecting new genes (Tung et al. 2010; Zhao et al. 2011; Cook et al. 2012; Xu et al. 2017a, b). Recently, the wheat 90 K iSelect SNP genotyping array was used to construct high-density linkage maps for QTL analysis in bread and durum wheats (Wang et al. 2014; Gao et al. 2015; Jin et al. 2016; Zhai et al. 2016; Wen et al. 2017). In addition, once genes and QTL have been cloned, gene-based SNP haplotypes can be identified to improve selection of target alleles in breeding programs (Rasheed et al. 2016).

The aims of the present study were to: (1) identify QTL and tightly linked SNP markers for yield-related traits using three RIL populations, (2) detect QTL clusters or pleiotropic loci associated with those traits, and (3) investigate genetic relationships among different yield-related traits.

Materials and methods

Plant materials and field trials

Three RIL populations derived from crosses Doumai × Shi 4185 (D × S, 275 F2:6 RILs), Gaocheng 8901 × Zhoumai 16 (G × Z, 176 F2:6 RILs) and Linmai 2 × Zhong 892 (L × Z, 273 F2:6 RILs) were used for constructing high-density linkage maps and QTL mapping. Doumai is a winter wheat line with a larger spike, TKW and flag leaf area and a lower tillering capacity; Shi 4185 (pedigree: Zhi 8094/Baofeng 7228//Shi 84-7120) is a winter wheat cultivar released in Hebei province, which has a smaller spike and TKW, and a higher tillering capacity. Gaocheng 8901 (pedigree: 77546-2/Linzhangmai) was cultivated in Hebei province with a taller plant height (above 90 cm), a smaller flag leaf width and good processing quality, while Zhoumai 16 (pedigree: Zhoumai 9/Zhou 8425B) is a widely grown facultative wheat cultivar in Henan province, which has a short stature (about 70 cm) and a larger spike and flag leaf width. Linmai 2 (pedigree: Lumai 23/Lin 90-15) is a semi-winter wheat cultivar in Shandong province with a larger spike and TKW and a lower tillering capacity, whereas Zhong 892 is a semi-winter wheat line from Henan province with a fewer KNS, a shorter stature (about 73 cm), and a narrow flag leaf.

The D × S population was grown at Shunyi in Beijing and Shijiazhuang in Hebei province during the 2012–2013, 2013–2014 and 2014–2015 cropping seasons; the G × Z and L × Z populations were grown at Anyang in Henan province and Suixi in Anhui province during the 2012–2013 and 2013–2014 cropping seasons. A randomized complete block design with three replicates was employed. Each plot comprised three 1.5 m rows spaced 20 cm apart, with 50 seeds in each row. Agronomic management was performed according to local practices at each location.

Phenotyping

Twelve phenotypic traits, SN, KNS, TKW, KL, KW, SL, SDW, HD, PH, UIL, FLL and FLW, were assessed in each population (Tables S1, S2 and S3).

Among yield components, SN was estimated by counting spikes at physiological maturity in one meter row section of each plot and expressed as m−2; KNS was measured by the mean kernel number of 20 randomly selected spikes in each plot at physiological maturity; and TKW was determined by the weight of 500 kernels from each plot after harvest. For kernel-related traits, 20 kernels were randomly selected from each plot after harvest for measuring KL and KW, represented by the mean length and width of 20 kernels, respectively. For spike-related traits, 20 spikes from 20 single plants in each plot were randomly selected at physiological maturity for measuring SL and SDW, the mean distance between the base and the top excluding the awns and the mean weight of 20 spikes after drying at 75 °C for 48 h, respectively. HD was recorded at half-spike emergence on a plot basis and represented as days from sowing to heading. For PH-related traits, 10 single plants in each plot were randomly selected at physiological maturity for measuring PH and UIL as the mean distance between the stem base and the top of spikes excluding awns and the mean length of the uppermost internode, respectively. For flag leaf area-related traits, 10 random flag leaves in each plot at the mid grain-fill stage were used to measure FLL and FLW, represented by the distance between the base and the tip, and width at the widest point, respectively.

Statistical analysis

For each trait, best linear unbiased estimation (BLUE) were calculated across environments using the ANOVA function in IciMapping V4.0 software (Li et al. 2007; http://www.isbreeding.net) assuming fixed effects for the genotype. Analysis of variance (ANOVA) was computed using the PROC GLM procedure in SAS 9.4 software (SAS Institute, http://www.sas.com). Broad-sense heritabilities (h2) were calculated following Nyquist and Baker (1991). PROC CORR procedure was used to calculate Pearson’s correlation coefficients (r) among the traits based on BLUE values.

Genotyping and high-density linkage map construction

The populations were genotyped using the wheat 90 K iSelect SNP array outsourced from Capital Bio Corporation, Beijing, China (http://www.capitalbiotech.com/). To reduce the influence of low-quality SNPs on the mapping results, three approaches were used to control the SNP data: (1) the heterozygous loci were treated as missing data, (2) the SNPs with missing data > 10% were filtered, and (3) allelic frequencies at a locus should be between 0.3 and 0.7.

The construction of high-density linkage maps was previously described by Wen et al. (2017) and the main steps are as follows: firstly, the BIN function of the IciMapping V4.0 software (Li et al. 2007; http://www.isbreeding.net) was used to place the SNP markers with no recombination into one bin and identify markers with minimum percentage of missing data as frame markers; secondly, the “Grouping” function in JoinMap 4.0 (Stam, 1993) was employed in sorting the frame markers into different groups with LOD thresholds ≥ 7; thirdly, the “AutoMap” command in MapDisto 1.7 (Lorieux, 2012) was used to order the frame markers and calculate genetic distances; finally, based on the CSS database (IWGSC RefSeq v1.0, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php), linkage groups were assigned and oriented on chromosomes.

QTL analysis

The frame markers identified in each bin were used for QTL mapping in IciMapping V4.0 software (Li et al. 2007; http://www.isbreeding.net). Composite interval mapping (ICIM) was chosen to search for QTL of phenotypic traits of all lines from each environment and BLUE values. LOD scores to declare significant QTL for all traits ranged from 2.0 to 2.5 across environments following 2000 permutations at P = 0.01; thus a LOD threshold of 2.5 was chosen for declaration of putative QTL. QTL detected in different environments for one trait were considered to be the same if the distance between the LOD contour peaks is less than 20 cM. QTL for same traits identified in more than half of the environments tested were considered to be stable.

QTL position comparison and candidate gene discovery

QTL identified in each population were integrated on the consensus map constructed in our previous study (Wen et al. 2017). The positions of QTL for the same traits in different populations were compared based on consensus maps (Wang et al. 2014; Wen et al. 2017). Flanking sequences of SNPs linked to QTL within a confidence interval (Darvasi and Soller 1997) were also used to blast against the CSS database (IWGSC RefSeq v1.0, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php) to identify the positions of QTL on the physical map. QTL for same traits mapped in different populations were considered to be the same if their positions were close in all three maps.

Candidate genes for QTL were determined using the following two approaches. Firstly, sequences of cloned genes on the same chromosome as QTL for each trait were used to blast against the CSS database. The gene was considered to be a candidate gene if its location was the same as SNPs mapped within a confidence interval to the LOD contour peak of the QTL on the physical map. Secondly, the flanking sequences of SNPs mapped within a confidence interval to the LOD contour peak of QTL were used to blast against the NCBI database (http://www.ncbi.nlm.nih.gov/) to identify putative gene functions. BLAST hits were filtered to an e-value threshold of 10−5 with sequence similarities higher than 75%.

Results

Phenotypic evaluation

The distributions of all 12 traits revealed continuous variation and transgressive segregation in the three populations, indicating polygenic inheritance. Among these, SN, KNS, TKW, SL, SDW, PH, UIL, FLL and FLW exhibited larger ranges in the populations, whereas KL, KW and HD showed less variation (Tables S1, S2 and S3).

ANOVA showed highly significant effects (P < 0.01) of lines and environments on all traits (Tables S8, S9 and S10). KNS, TKW, KL, KW, SL, PH and FLW had high h2 (above 0.80) in all three populations, whereas SN, SDW, HD, UIL and FLL showed moderate h2 (0.60–0.80), indicating that most of these traits were stable and mainly determined by genetic factors (Tables S8, S9 and S10).

SN exhibited significant (P < 0.01) and negative correlations with KNS, TKW, KW, SDW and FLW in both the D × S and L × Z populations (Tables S11, S12 and S13). KNS was significantly and positively correlated with SDW and FLW in all three populations (Tables S11, S12 and S13). TKW had significant and positive correlations with KL, KW and PH in all three populations, and the highest correlation was between TKW and KW (r = 0.71–0.82) in all three populations (Tables S11, S12 and S13).

High-density linkage maps

Based on the criteria mentioned above, 11,012 (D × S), 11,979 (G × Z) and 10,443 (L × Z) markers were used for the construction of genetic linkage maps (Wen et al. 2017). For the D × S population, all polymorphic markers were assigned to 2851 recombinant bins, and the total length of the linkage map constructed by the frame markers is 2030.0 cM, with an average frame marker density of 1.40 markers per cM (Tables S4 and S7). For the G × Z population, all polymorphic markers were placed into 3284 recombinant bins, covering a total length of 3130.3 cM, with an average frame marker density of 1.04 markers per cM (Tables S5 and S7). For the L × Z population, all polymorphic markers were sorted into 3489 recombinant bins, and the total length of the linkage map constructed by the frame markers is 2954.8 cM, with an average frame marker density of 1.08 markers per cM (Tables S6 and S7).

QTL for yield-related traits

Sixty five, 46 and 53 QTL for yield-related traits were identified in the D × S, G × Z and L × Z populations, respectively (Table 1 and Fig. 1).

Table 1 QTL identified for yield-related traits in Doumai × Shi 4185, Gaocheng 8901 × Zhoumai 16, and Linmai 2 × Zhong 892 RIL populations
Full size table
Fig. 1
figure 1figure 1figure 1figure 1

QTL for yield-related traits on the wheat consensus map (Wen et al. 2017). Markers are indicated on the right side, and corresponding centimorgan (cM) distances are shown on the left. QTL confidence intervals with a LOD score ≥ 2.5 are indicated by vertical bars. Different shapes and colors represent QTL identified in different populations and traits, respectively

Full size image

QTL for yield components

Nine QTL for SN were mapped on chromosomes 1AL, 1BL, 2BS, 3AL, 4BS (2), 5DL, 6AL and 7BS, with 4, 3 and 2 in the D × S, G × Z and L × Z populations, respectively (Table 1). QSN.caas-4BS was stably identified in the D × S and L × Z populations, explaining 22.0–59.0 and 16.0–23.3% of the phenotypic variances across environments, respectively (Tables 1, 3). QSN.caas-1AL (closest marker IWB40317) identified in two environments and BLUE value in the G × Z population explained 9.1–16.2% of the phenotypic variance.

Sixteen QTL for KNS were identified on chromosomes 1BS, 1BL (2), 3DL, 4AL, 4BS (3), 5BL, 5DL, 7AS, 7AL (2), 7BS, 7BL and 7DS, with 5, 7 and 4 in the D × S, G × Z and L × Z populations, respectively (Table 1). QKNS.caas-4BS was stably detected in the D × S and L × Z populations, accounting for 9.9–43.1 and 17.8–36.6% of the phenotypic variances, respectively (Tables 1, 3). QKNS.caas-7AL was also stably identified in the D × S and L × Z populations, explaining 3.5–4.7 and 2.7–5.2% of the phenotypic variances, respectively. QKNS.caas-3DL (closest marker IWB66510) in the G × Z population was significant in two environments, accounting for 8.0–10.8% of the phenotypic variance.

Seventeen QTL for TKW were detected on chromosomes 2AS, 2BS, 3AL, 3B, 3DL, 4AL, 4BS (2), 4DS (2), 5AL (2), 5DL, 6AL, 6BL, 7AL and 7BL, with 8, 3 and 6 in the D × S, G × Z and L × Z populations, respectively (Table 1). QTKW.caas-4BS was stably identified in the D × S and G × Z populations, explaining 12.1–45.6 and 6.6–9.2% of the phenotypic variances, respectively (Tables 1, 3). QTKW.caas-5AL, QTKW.caas-5DL and QTKW.caas-6BL in the D × S population and QTKW.caas-3B and QTKW.caas-7AL.1 in the L × Z population were detected in most environments and BLUE values. QTKW.caas-6AL (closest marker IWA428) in the L × Z population was found in two environments and BLUE value, accounting for 10.3–18.8% of the phenotypic variance.

QTL for kernel-related traits

Eleven QTL for KL were mapped on chromosomes 1AL, 1BS, 1BL, 4BL, 4DS, 5BS, 5DL, 6AL, 6DL, 7AS and 7DS, with 6, 3 and 2 in the D × S, G × Z and L × Z populations, respectively (Table 1). QKL.caas-5BS and QKL.caas-5DL in the D × S population and QKL.caas-1BL and QKL.caas-6DL in the L × Z population were stably identified in all environments and BLUE values. QKL.caas-1BS (closest marker IWB31844) in the G × Z population was significant in three environments and BLUE value, explaining 8.2–17.1% of the phenotypic variance.

Thirteen QTL for KW were identified on chromosomes 2BL, 3AL, 3B (2), 4AS, 4AL, 4BS, 4DS, 5AL, 5BL, 5DL, 6AL and 7AL, with 4, 3 and 6 in the D × S, G × Z and L × Z populations, respectively (Table 1). QKW.caas-4AL and QKW.caas-4BS in the D × S population, QKW.caas-3AL and QKW.caas-5AL in the G × Z population and QKW.caas-2BL, QKW.caas-3B.2 and QKW.caas-4AS in the L × Z population were stably identified in most of the environments and BLUE values. QKW.caas-3B.1 explained 8.4–13.9% of the phenotypic variance in the G × Z population.

QTL for spike-related traits

Seventeen QTL for SL were detected on chromosomes 1AL, 2DS (3), 2DL, 3B, 3DL, 4AL, 4BS, 5AL, 6AL (2), 6BL (4) and 6DL, with 6, 6 and 5 in the D × S, G × Z and L × Z populations, respectively (Table 1). QSL.caas-2DS stably detected in the G × Z and L × Z populations accounting for 21.1–29.9 and 3.2–8.3% of the phenotypic variances, respectively (Tables 1, 3); QSL.caas-6AL was stably identified in the D × S and L × Z populations, explaining 1.7–3.6 and 2.7–4.8% of the phenotypic variances, respectively; QSL.caas-6BL.1 was stably detected in all three populations, accounting for 9.8–28.5, 4.8–9.2 and 10.4–19.1% of the phenotypic variances, respectively. QSL.caas-2DS.1, QSL.caas-4AL, QSL.caas-5AL.2 in the G × Z population, QSL.caas-2DL in the D × S population and QSL.caas-4BS in the L × Z population were stably identified in four or five environments and BLUE values.

Ten QTL for SDW were identified on chromosomes 1BS, 4BS (2), 5DL (2), 6AL (2), 6BL, 7AS and 7BL, with 4, 3 and 3 in the D × S, G × Z and L × Z populations, respectively. QSDW.caas-4BS was stably detected in the D × S and L × Z populations, explaining 37.2–42.9 and 6.4–19.3% of the phenotypic variances, respectively (Tables 1, 3). QSDW.caas-6BL (closest marker IWB73837) in the D × S population was significant in two environments and BLUE value, accounting for 15.2–28.3% of the phenotypic variance; QSDW.caas-5DL.2 (closest marker IWB12546) in the D × S population was stably detected in all four environments and BLUE value.

QTL for HD

Nineteen QTL for HD were mapped on chromosomes 1BL, 1DL (2), 2AL, 2BS (2), 2DS, 3AS, 3AL (2), 4BS, 5AL, 5DL (2), 6BL (2), 7BS, 7BL and 7DL, with 6, 6 and 7 in the D × S, G × Z and L × Z populations, respectively (Table 1). Three common QTL were detected on chromosomes 2BS, 5DL and 6BL in the D × S and L × Z populations, accounting for 5.1–16.0, 3.0–16.5 and 2.7–11.0% of the phenotypic variances, respectively (Tables 1, 3). QHD.caas-1DL.2 (closest marker IWB49684) was a major QTL and stably identified in all environments and BLUE value in the G × Z population.

QTL for PH-related traits

Eighteen QTL for PH were identified on chromosomes 1BL, 2AL, 2BL (2), 3AS, 3AL, 3B, 4BS (2), 4BL, 4DS (2), 5AL (2), 5BL, 6BL (2) and 6DS, with 9, 5 and 4 in the D × S, G × Z and L × Z populations, respectively (Table 1). QPH.caas-4BS was stably identified in the G × Z and L × Z populations, accounting for 13.2–21.1 and 7.4–23.7% of the phenotypic variances, respectively; QPH.caas-4DS was also stably identified in the G × Z and L × Z populations, explaining 9.5–15.0 and 6.7–18.5% of the phenotypic variances, respectively; QPH.caas-6BL was stably detected in the D × S and L × Z populations, accounting for 2.3–4.8 and 3.4–11.9% of the phenotypic variances, respectively (Tables 1, 3). QPH.caas-1BL, QPH.caas-5AL.1 and QPH.caas-6DS in the D × S population and QPH.caas-2BL.2 in the L × Z population were stably identified in all environments and BLUE values.

Fifteen QTL for UIL were detected on chromosomes 2AS (3), 2AL, 2BL, 3AL, 3DL, 4AS, 4BS (3), 4DS, 5AL, 6BL and 7DS, with 6, 4 and 5 in the D × S, G × Z and L × Z populations, respectively (Table 1). QUIL.caas-2AS was stably identified in the D × S and L × Z populations, accounting for 8.5–10.8 and 2.0–3.2% of the phenotypic variances, respectively; QUIL.caas-4BS stably detected in all three populations, explained 4.0–7.0, 7.1–17.2 and 7.0–10.7% of the phenotypic variances, respectively (Tables 1, 3). QUIL.caas-4DS and QUIL.caas-6BL were significant and stable across environments, explaining 7.9–18.6 and 10.4–13.8% of the phenotypic variances in the G × Z and L × Z populations, respectively. QUIL.caas-3AL and QUIL.caas-4AS in the L × Z population, and QUIL.caas-3DL in the D × S population were stably identified in all environments and BLUE values.

QTL for flag leaf-related traits

Six QTL for FLL were mapped on chromosomes 2AL, 2BL, 3B, 4AS, 4BS and 5DL, with 3 and 3 in the D × S and L × Z populations, respectively (Table 1). QFLL.caas-3B (closest marker IWB54417) was significant in two environments and BLUE value in the L × Z population, explaining 10.1–14.1% of the phenotypic variances.

Thirteen QTL for FLW were identified on chromosomes 1BL, 2BS (2), 4AL (2), 4BS (2), 5AL (2), 5BL, 5DL, 6AS and 6BL, with 4, 3 and 6 in the D × S, G × Z and L × Z populations, respectively (Table 1). QFLW.caas-2BS was stably detected in the G × Z and L × Z populations, accounting for 10.5–14.3 and 1.9–4.8% of the phenotypic variances, respectively; QFLW.caas-4BS, stably identified in the D × S and L × Z populations, explained 2.8–15.5 and 11.5–22.1% of the phenotypic variances, respectively (Tables 1, 3). QFLW.caas-4AL.2 (closest marker IWB38036) in the G × Z population was significant in two environments and BLUE value. QFLW.caas-5BL and QFLW.caas-5DL in the D × S population, and QFLW.caas-5AL.1, QFLW.caas-6AS and QFLW.caas-6BL in the L × Z population were stably identified in all environments and BLUE values.

QTL clusters and pleiotropic loci

Eleven QTL clusters were identified on chromosomes 1BL, 2AS, 3DL, 4BS, 4DS, 5AL (2), 5DL, 6AL (2) and 6BL (Table 2). Clusters on chromosomes 4BS, 4DS and 6BL were observed in all three populations, whereas clusters on chromosomes 5DL and 6AL (interval 74.3–75.9 cM) were identified in the D × S and L × Z populations. Among these QTL clusters, 10 were associated with kernel-related traits, except for that on chromosome 3DL. The interval 15.5–32.3 cM on chromosome 4BS harboring 20 QTL for all traits except for KL, with 7, 4 and 9 QTL in the D × S, G × Z and L × Z populations, respectively; intervals 32.7–45.1 cM and 17.0–30.3 cM on chromosomes 2AS and 4DS, respectively, were associated with both kernel and PH-related traits; intervals 12.7–50.4 cM and 86.6–105.7 cM on chromosomes 5DL and 6BL harboring 11 and 9 QTL, respectively, were important in determining yield-related traits.

Table 2 QTL clusters for yield-related traits identified in Doumai × Shi 4185, Gaocheng 8901 × Zhoumai 16, and Linmai 2 × Zhong 892 RIL populations and their positions on the consensus map (Wen et al. 2017)
Full size table

Discussion

Comparison with previous studies

The Green Revolution genes Rht-B1b and Rht-D1b greatly contributed to wheat yield potential through improving lodging resistance and modifying dry matter distribution (Peng et al. 1999; Ellis et al. 2005). Blasting results indicated that the QTL for PH on chromosomes 4BS and 4DS in the G × Z and L × Z populations were Rht-B1b and Rht-D1b, respectively (Peng et al. 1999). These two genes were also detected by Gao et al. (2015) and Sun et al. (2017) using the wheat 90 K iSelect SNP array. Interestingly, we identified QTL for SN, KNS, TKW, KW, SDW, UIL and FLW around the Rht-B1 locus on chromosome 4BS with large contributions in the D × S population, whereas no QTL for PH was detected. Sequencing of this locus indicated that there is a large deletion in Doumai, whereas the allele in Shi 4185 is Rht-B1b (data not shown). Four QTL for KNS, TKW, PH and UIL in the G × Z population and nine QTL for SN, KNS, SL, SDW, HD, PH, UIL, FLL and FLW in the L × Z population were detected around the Rht-B1 locus. These were attributed to the effect of Rht-B1, in agreement with previous reports (Cuthbert et al. 2008; Deng et al. 2011; Zhang et al. 2017b). Gao et al. (2015) identified QTL for TKW and KW at the Rht-B1 locus, and QTL were similarly reported by Gegas et al. (2010) and Ramya et al. (2010).

Six QTL for PH and kernel-related traits around the Rht-D1 locus were detected in the G × Z and L × Z populations, indicating a lesser effect on yield-related traits than Rht-B1. Gao et al. (2015) identified four QTL for GY, chlorophyll content and normalized difference in vegetation index around Rht-D1. Cuthbert et al. (2008) and Zhang et al. (2017b) both reported a QTL-rich region for yield-related traits around the Rht-B1 locus, whereas no QTL-rich regions around Rht-D1 were identified in these studies.

Many studies to date have implied that SN is controlled by polygenes (Cuthbert et al. 2008; Cui et al. 2014; Lee et al. 2014; Gao et al. 2015). A minor QTL on chromosome 6AL identified in the present study was also detected by Gao et al. (2015). Another QTL on chromosome 3AL is at a similar position with that in Lee et al. (2014). The major QTL on chromosome 1AL detected in two environments and BLUE value is likely to be new.

QKNS.caas-7AL, stably identified in the D × S and L × Z populations, is at the same position as SNP wsnp_Ex_c11047_17915103 detected by Sun et al. (2017). Another minor QTL on chromosome 7BS was previously observed by Gao et al. (2015). QKNS.caas-1BL.2 and QKNS.caas-7BL are close to those in Xu et al. (2017a, b) and Wang et al. (2011a), respectively. The major QTL on chromosome 3DL is likely to be new.

TKW-associated QTL were mapped on all 21 chromosomes in previous studies (Gegas et al. 2010; Ramya et al. 2010; Gao et al. 2015; Wu et al. 2015; Sukumaran et al. 2015; Cheng et al. 2017; Sun et al. 2017). A gene for TKW on chromosome 7D was finely mapped by Röder et al. (2008) with the SSR marker Xgwm1002-7D. Blasting results revealed that QTKW.caas-3B and QTKW.caas-6AL are probably TaGS5 and TaTPP-6AL1, respectively (Wang et al. 2015; Ma et al. 2016; Wang et al. 2016b; Zhang et al. 2017c), whereas QTKW.caas-2BS, QTKW.caas-3DL.1 and QTKW.caas-7AL.1 are different from the cloned genes TaSus2-2B (2BS), TaCKX6-D1 (3D) and TaGASR7-A1 (7AL) (Jiang et al. 2011; Zhang et al. 2012b; Dong et al. 2014), respectively. The 3AL QTL is at the same position as SNP Excalibur_c39508_88 associated with TKW (Sun et al. 2017). In addition, QTKW.caas-2AS, QTKW.caas-4AL.1, QTKW.caas-4DS.1 and QTKW.caas-6BL are located at similar positions to those in Cui et al. (2014), Wu et al. (2015), Mohler et al. (2016) and Mir et al. (2012), respectively. QTKW.caas-5AL, QTKW.caas-5DL and QTKW.caas-7AL.1 are likely to be new QTL detected in most environments and BLUE values.

Blasting results indicated that QKL.caas-6DL is probably TaGS1a (Guo et al. 2013); QKL.caas-1BL and QKL.caas-5BS are at the same positions as those in Wu et al. (2015) and Zhai et al. (2018), respectively. QKL.caas-4DS, co-localized with QTKW.caas-4DS, is at similar position to that in Gegas et al. (2010). In addition, QKL.caas-1AL, QKL.caas-1BS, QKL.caas-6AL, QKL.caas-7AS and QKL.caas-7DS are at similar positions to those in Cui et al. (2014), Mohler et al. (2016), Li et al. (2015), Tyagi et al. (2014) and Gegas et al. (2010), respectively. QKL.caas-5DL, co-localized with QTKW.caas-5DL and stably identified in all six environments and BLUE value, is likely to be a new QTL.

QKW.caas-6AL, co-localized with QTKW.caas-6AL, is likely to be TaTPP-6AL1 (Zhang et al. 2017c); QKW.caas-3B.2, identified in all four environments and BLUE value, is at the same position as that in Zhai et al. (2018). QKW.caas-3AL, co-localized with QTKW.caas-3AL, is at a similar position to SNP Excalibur_c39508_88 for TKW (Sun et al. 2017). QKW.caas-5BL and QKW.caas-5DL are at similar positions to those in Cui et al. (2014) and Mohler et al. (2016), respectively. QKW.caas-5AL, co-localized with QTKW.caas-5AL.3, is likely to be a new QTL. QKW.caas-2BL, QKW.caas-3B.1, QKW.caas-3B.2, QKW.caas-4AS and QKW.caas-4AL are also new QTL.

QSL.caas-6BL.1, identified in all three populations, is at the same position as those in Gao et al. (2015) and Sun et al. (2017); QSL.caas-2DS, detected in the G × Z and L × Z populations, is at a similar position to that in Azadi et al. (2015); QSL.caas-6AL, identified in the D × S and L × Z populations, is at a similar position to QTL reported previously (Lee et al. 2014; Liu et al. 2014). QSL.caas-3B, QSL.caas-3DL, QSL.caas-4AL and QSL.caas-5AL.2 are also in similar positions to those reported in Azadi et al. (2015), Liu et al. (2014), Gao et al. (2015) and Wang et al. (2011a), respectively. QSL.caas-2DS.1 and QSL.caas-2DL identified in four or five environments and BLUE values are likely to be new.

QSDW.caas-6BL and QSDW.caas-7BL are at similar positions as SNPs RAC875_c31299_1302 and BS00055584_51 identified by Valluru et al. (2017), respectively. QSDW.caas-4BS stably identified in the D × S and L × Z populations is likely to be a new QTL. QSDW.caas-5DL detected in all four environments and BLUE value is also likely to be new.

Blasting results showed that QHD.caas-5DL in the D × S and L × Z populations should be contributed by Vrn-D1 (Fu et al. 2005); QHD.caas-7BS is about 6 Mb from Vrn-B3 locus (Yan et al. 2006) on the physical map (IWGSC RefSeq v1.0, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php); QHD.caas-2BS, identified in the D × S and L × Z populations and close to the Eps QTL detected by Le Gouis et al. (2012), is about 56 Mb from Ppd-B1 locus (Nishida et al. 2013); QHD.caas-2DS is about 46 Mb from Ppd-D1 (Beales et al. 2007). Because there are no variations at Vrn-B3 and Ppd-D1 loci in the parents of three populations, QHD.caas-7BS and QHD.caas-2DS should be different from Vrn-B3 and Ppd-D1, respectively. Comparison of the physical locations of QHD.caas-2BS and Ppd-B1 also shows that the two loci are different. Previously, Wang et al. (2016a) cloned gene TaELF3-1DL associated with early flowering on chromosome 1DL using the G × Z population. This locus is consistent with QHD.caas-1DL.2 in the present study and at the same position as the major Eps QTL flanked by Xgdm111 and Xbarc62 (Griffiths et al. 2009; Zikhali et al. 2014). Another QTL on chromosome 1DL (QHD.caas-1DL.1) about 60 Mb from QHD.caas-1DL.2 should be the flowering time-related gene TaFT3-D1, which is a homolog of the barley gene HvFT3 (Faure et al. 2007; Halliwell et al. 2016). QHD.caas-1BL, QHD.caas-3AL.1, QHD.caas-5AL, QHD.caas-7BL and QHD.caas-7DL are at similar positions to those in Zhang et al. (2009), Edae et al. (2014), Reif et al. (2011), Bordes et al. (2013) and Cuthbert et al. (2008). QHD.caas-6BL identified in the D × S and L × Z populations is probably a new QTL.

Rht4 and Rht12 were identified as induced semi-dwarf mutants in wheat and mapped on chromosomes 2BL and 5AL, respectively (Konzak 1988; Ellis et al. 2005). Blasting results indicated that QPH.caas-2BL.1 is about 7 Mb from marker Xwmc317 linked to Rht4 on the physical map (IWGSC RefSeq v1.0, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php). Nevertheless, no sufficient evidence can prove that they are the same. Two QTL for PH on chromosome 5AL are different from Rht12, whereas QPH.caas-5AL.1 is at a similar position to the QTL detected by Zhang et al. (2017b). QPH.caas-1BL, QPH.caas-2BL.2, QPH.caas-6BL and QPH.caas-6DS are likely to be new.

UIL plays an important role in determining PH and GY, but relatively little attention was given to this trait in previous studies. Our results indicated that Rht-B1 and Rht-D1 affect UIL based on the co-localization of PH and UIL QTL at two loci. QUIL.caas-2AS, identified in the D × S and L × Z populations, was located at the same position as a gene Rht_NM9 that was finely mapped within a region of 8.86 Mb on chromosome 2AS (Lu et al. 2015), and close to the QTL in Yu et al. (2014). Zhang et al. (2017b) reported a QTL associated with PH and related traits close to QUIL.caas-3DL. The major QTL for UIL on chromosome 6BL, with no correlation with PH, is likely to be new.

Genetic studies on flag leaf-related traits (FLL and FLW) in wheat have always been a big challenge due to instability across environments. In rice, two genes (Nal1 and Nal7) for FLW have been cloned, and two QTL (qFL1 and qFLW4) associated with FLL and FLW, respectively, were finely mapped (Fujino et al. 2008; Qi et al. 2008; Wang et al. 2011b; Chen et al. 2012). However, to date, no genes for FLL and FLW have been cloned in wheat, and one gene (TaFLW1) next to Fhb5 on chromosome 5A was finely mapped (Xue et al. 2013). QFLL.caas-2BL, previously identified by Wu et al. (2016), was close to a QTL reported in Edae et al. (2014). QFLL.caas-2AL and QFLL.caas-3B detected in two or three environments and BLUE values are probably new.

Several previous studies indicated that chromosome 5A is important in determining FLW in wheat (Jia et al. 2013; Xue et al. 2013; Wu et al. 2016). Two QTL on chromosome 5AL were detected in the present study, and QFLW.caas-5AL.2 is at the same position as that in Wu et al. (2016). QFLW.caas-1BL and QFLW.caas-6BL are also at the same positions with those in Wu et al. (2016). QFLW.caas-2BS, QFLW.caas-4AL.2, QFLW.caas-5AL.1, QFLW.caas-5BL, QFLW.caas-5DL and QFLW.caas-6AS are likely to be new QTL.

Genetic relationships among grain yield-related traits

GY is a complex quantitative trait controlled by multiple genes and highly influenced by environment. As expected, yield-related traits are significantly correlated. In the present study, 11 QTL clusters or QTL-rich regions on chromosomes 1BL, 2AS, 3DL, 4BS, 4DS, 5AL (2), 5DL, 6AL (2) and 6BL were identified. The pleiotropic loci on chromosomes 4BS and 5DL were associated with 11 and 9 traits, respectively, showing crucial roles in affecting GY. The interval 15.5–32.3 cM around Rht-B1 locus on chromosome 4BS, harboring 20 QTL for all traits except KL, is a pleiotropic locus. This represented a strong validation that Rht-B1 has a significant influence on yield-related traits. Several studies have reported this QTL-rich region for yield (McCartney et al. 2005; Cuthbert et al. 2008; Zhang et al. 2017b) and physiological-related traits (Gao et al. 2015), and our study revealed more detail on the effect of this locus. Another QTL-rich region is around the Rht-D1 locus on chromosome 4DS, including QTL for TKW, KL, KW, PH and UIL. The large effect of Rht-D1 on TKW is consistent with the significantly positive correlation between TKW and PH-related traits. The QTL-rich region around this locus for leaf area (Wu et al. 2016) and physiological-related traits (Gao et al. 2015) were also identified previously.

Among the 11 QTL clusters, 10 were associated with kernel-related traits, indicating that the variation of kernel-related traits is easily caused by the altered agronomic traits. Six QTL for KW were co-localized with those for TKW, whereas only two for KL were co-localized with the TKW QTL, which was reflected by the higher correlations between KW and TKW (r = 0.71–0.82). Three QTL for SN detected among the pleiotropic regions were co-localized with TKW and KW QTL. As SN was significantly and negatively correlated with TKW in the three populations, the effects of these pleiotropic regions on SN and TKW are likely to be contrasting. In addition, QTL for SN and KNS on chromosomes 4BS and 5DL identified in the D × S population were co-localized, in agreement with the significant and negative correlation between two traits (r = − 0.66). Three KNS QTL were co-localized with PH or UIL QTL, and three with kernel-related traits. Due to the negative correlations, increased PH, UIL and kernel-related traits may have adverse effects on KNS. Moreover, four QTL for SDW on chromosomes 1BS, 4BS, 5DL and 7AS and four for FLW on chromosomes 1BL, 4AL, 4BS and 5DL were co-localized with the QTL for KNS, in agreement with the significant correlations among these traits. The HD QTL were co-localized with those for TKW, SL, SDW and PH, respectively. Based on the correlations between HD and each trait, it was considered that taller plants are likely to be heading earlier, whereas the earlier heading date is favorable for spike development and grain-filling. Le Gouis et al. (2012) identified two HD loci close to Rht-B1 and Rht-D1, but the conclusion on effects of two genes for HD remains controversial.

Molecular mechanisms underlying yield-related traits

With availability of the CSS database (IWGSC RefSeq v1.0, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php) QTL mapped with high-density SNPs can be in comparison with cloned genes. Compared to SSR markers, the 90 K iSelect SNP assay was developed from the transcriptome (Wang et al. 2014), providing a better approach in searching for candidate genes. In the present study, 11 candidate genes were identified, including eight cloned genes for kernel traits, HD and PH as well as three predicted genes for TKW, SL and HD (Table S14).

GA-insensitive plants carrying the Rht-B1b, Rht-B1e and Rht-D1b alleles were characterized by reduced height, because of decreased sensitivity of their vegetative tissues to endogenous gibberellin (GA) (Keyes et al. 1989). This effect is manifested by decreased cell length in almost all vegetative organs, followed not only by reduced plant height, but also by decreased coleoptile, internode lengths, and leaf area. Chebotar et al. (2016) reported that the GA-sensitive (Rht8) and GA-insensitive (Rht-B1 and Rht-D1) dwarfing genes had pleiotropic effects on all traits studied except for the number of fertile spikelets. In the present study, Rht-B1 and Rht-D1 as well as several QTL for PH were identified. As expected, Rht-B1 and Rht-D1 showed large effects on PH and yield-related traits, consistent with the previous reports (Peng et al. 1999; Ellis et al. 2005; Gao et al. 2015; Sun et al. 2017). Moreover, other QTL for PH were always co-localized with the QTL for yield traits, especially for kernel-related traits, indicating similar but lesser effects compared to Rht-B1 and Rht-D1.

Kernel-related genes are involved in the regulation of cell division and expansion, presumably via three major pathways, including ubiquitination-mediated proteasomal degradation, G-protein signaling, and phytohormones, and other unknown pathways (Huang et al. 2009; Li et al. 2011; Ma et al. 2016; Zhang et al. 2017c). Three QTL for kernel-related traits in the present study were associated with cloned genes TaGS5 (3BS), TaTPP-6AL1 (6AL) and TaGS1a (6DL) (Guo et al. 2013; Ma et al. 2016; Zhang et al. 2017c), respectively. Moreover, the co-localized QTL (QTKW.caas-5AL.3 and QKW.caas-5AL) putatively correspond to the isoamylase 3 gene EMS47868.1 involved in biosynthesis of amylopectin in wheat and rice. The function of this gene is similar to Trehalose 6-phosphate phosphatase (TPP), which is associated with starch accumulation and grain yield in several crops (Zhang et al. 2017c).

The effect of Ppd and Vrn genes on HD or flowering time is relatively well understood; however, much less is known about the influence of genetic factors that modify flowering time once photoperiod and vernalization requirements have been met. Three QTL for HD in the present study are associated with cloned genes Vrn-D1, TaFT3-D1 and TaELF3-1DL (Fu et al. 2005; Faure et al. 2007; Halliwell et al. 2016; Wang et al. 2016a), respectively. Another QTL for HD on chromosome 4BS corresponds to Tubulin/FtsZ family protein gene NP_849388.1 involved in plant growth rhythm. In addition, a putative candidate gene corresponding to the bidirectional sugar transporter SWEET6b (EMS56832.1) is associated with SL on chromosome 2DS. The molecular mechanism for many yield-related traits is still not well understood, and more experimental analysis should be undertaken to confirm the roles of these genes on yield-related traits.

Potential implications in wheat breeding

Wheat GY is significantly influenced by environment, which creates major challenges in selection of high-yielding lines in small plots at the early stages of breeding programs. In contrast, GY components, as well as kernel, spike, PH and flag leaf-related traits are less influenced by environment. Consequently, more attention has focused on yield-related traits so as to improve GY. It has been proven previously, and in the present study, that TKW is significantly affected by KW (Huang et al. 2015; Wu et al. 2015). Considering the problems of kernel color preference and possible existence of black point, TKW can be selected directly in wheat breeding.

The present study showed that KNS was significantly and positively correlated with SDW and FLW in wheat. This was supported by the co-localization of QTL for these three traits. Due to its simplicity FLW can be used to select lines with large KNS. More physiological traits should be considered in future studies to dissect factors associated with SN.

The presence of common QTL identified in two or three populations reveals that the QTL have consistently stable effects in different genetic backgrounds. Therefore, the common QTL listed in Table 3 might be valuable to be employed in improving GY. In addition, stable QTL with large contributions identified in a single population, such as QTKW.caas-5DL, QKL.caas-5BS, QKL.caas-5DL, QKL.caas-6DL, QKW.caas-3B.2, QSL.caas-4AL, QPH.caas-5AL.1 and QFLW.caas-6AS, are also useful for MAS in breeding varieties adapted to specific environments. Moreover, flanking SNP markers of pleiotropic loci or QTL-rich regions for yield-related traits are suitable to select varieties with good performance. Our study proved that QTL mapping using three RIL populations together with high-density linkage maps offered an opportunity to identify common and stable QTL for MAS.

Table 3 Common QTL identified across different populations and their positions on the consensus maps (Wang et al. 2014; Wen et al. 2017) and (IWGSC RefSeq v1.0, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php), and comparisons with previous studies
Full size table

Kompetitive allele-specific PCR (KASP) is a homogenous, fluorescence-based genotyping variation of PCR (Rasheed et al. 2016). Currently, KASP assays are widely used in genotyping important traits due to high-throughput and low cost. Semi-thermal asymmetric reverse PCR (STARP) is a newly developed genotyping method adapting to multiple platforms and throughputs (Long et al. 2017). The development of these two assays offers better choices in detecting InDels or SNPs. SNP markers tightly linked with yield-related traits detected by QTL mapping or GWAS can be successfully transformed into KASP or STARP assays for MAS in wheat breeding (Rasheed et al. 2016; Liu et al. 2017). With the use of high-density linkage maps in gene mining and high-throughput platforms in genotyping, wheat breeding by molecular design is not a distant goal.