High-resolution detection of quantitative trait loci for seven important yield-related traits in wheat (Triticum aestivum L.) using a high-density SLAF-seq genetic map

Yield-related traits including thousand grain weight (TGW), grain number per spike (GNS), grain width (GW), grain length (GL), plant height (PH), spike length (SL), and spikelet number per spike (SNS) are greatly associated with grain yield of wheat (Triticum aestivum L.). To detect quantitative trait loci (QTL) associated with them, 193 recombinant inbred lines derived from two elite winter wheat varieties Chuanmai42 and Chuanmai39 were employed to perform QTL mapping in six/eight environments. A total of 30 QTLs on chromosomes 1A, 1B, 1D, 2A, 2B, 2D, 3A, 4A, 5A, 5B, 6A, 6D, 7A, 7B and 7D were identified. Among them, six major QTLs QTgw.cib-6A.1, QTgw.cib-6A.2, QGw.cib-6A, QGl.cib-3A, QGl.cib-6A, and QSl.cib-2D explaining 5.96-23.75% of the phenotypic variance were detected in multi-environments and showed strong and stable effects on corresponding traits. Three QTL clusters on chromosomes 2D and 6A containing 10 QTLs were also detected, which showed significant pleiotropic effects on multiple traits. Additionally, three Kompetitive Allele Specific PCR (KASP) markers linked with five of these major QTLs were developed. Candidate genes of QTgw.cib-6A.1/QGl.cib-6A and QGl.cib-3A were analyzed based on the spatiotemporal expression patterns, gene annotation, and orthologous search. Six major QTLs for TGW, GL, GW and SL were detected. Three KASP markers linked with five of these major QTLs were developed. These QTLs and KASP markers will be useful for elucidating the genetic architecture of grain yield and developing new wheat varieties with high and stable yield in wheat.

of high-yield wheat varieties to ensure future global food and nutrition security is an important target of the modern wheat breeding programs [2].
Wheat yield is a complex quantitative trait controlled by multiple genes and significantly influenced by interacting genetic and environmental factors [3,4]. By contrast, yield components including thousand grain weight (TGW), grain number per spike (GNS), grain width (GW), grain length (GL), plant height (PH), spike length (SL) and spikelet number per spike (SNS) typically show higher heritability than that of the yield [5][6][7]. Therefore, targeting these traits and identifying the related genes or quantitative trait loci (QTL) is an important approach to improve grain yield potential in wheat.
The molecular cloning of genes associated with wheat yield is difficult owing to wheat's huge and complicated genome. To date, only a few genes associated with grain yield have been cloned in wheat. For example, the application of semi-dwarfing genes Rht-B1b and Rht-D1b not only effectively improve the lodging resistance but also improve the harvest index, resulting in increasing yield since the 1970s [8][9][10]. The vernalization insensitive alleles of Vrn-1 (Vrn-A1, Vrn-B1, and Vrn-D1) shorten both the vegetative and the reproductive stages and have considerable impact on spike morphological traits [11,12]. The grain-shape gene TasgD1 encoding a Ser/Thr protein kinase glycogen synthase kinase3 and independently control semispherical grain trait [13]. A jasmonic acid synthetic gene keto-acyl thiolase 2B was cloned in a TGW mutant, showing significant effects on TGW and GW [14]. Additionally, homologous cloning is an effective approach to characterize gene in wheat. As of today more than 20 genes related to yield have been isolated through homologous cloning approach, including WFZP, WAPO1, TaGW7, TaGW2, TaCKX6-D1, TaTGW6, TaGASR7, TaGL3 and TaGS-D1 et al [15][16][17][18][19][20][21][22][23].
With the development of high-throughput sequencing technology, Single nucleotide polymorphisms (SNP) markers have been widely applied to construct high-density genetic maps for QTL mapping, due to their extensive and intensive distribution throughout genomes in many crop s [3,[36][37][38]. Specific-locus amplified fragment sequencing (SLAF-seq) was developed for economic and efficient high-throughput SNP discovery through restriction-site associated DNA tag sequencing (RAD-seq), which can provide abundant InDel and SNP markers to construct high-density genetic map [39][40][41].
In the present study, a high-resolution genetic map was constructed in a recombinant inbred line (RIL) population derived from two elite winter wheat varieties Chuanmai42 (CM42) and Chuanmai39 (CM39) based on SLAF-seq (Table S1, S2) [42]. Seven traits including TGW, GW, GL, PH, GNS, SL and SNS were assessed in multi-environments to detect potential major and stable QTL, which will lay out a foundation for further study on fine mapping and cloning of the underlying key genes for wheat yield.

Phenotypic variation
The phenotypic analysis showed that CM42 had higher trait values for TGW, GW, GL, GNS, PH and SL than those of CM39 in each of environments and the best linear unbiased prediction (BLUP) datasets (Table 1). In the RIL population, seven yield-related traits showed wide and significant variations in all environments and the BLUP datasets (Table 1). Of them, the TGW ranged from 20.81 to 72.7 gram (g), the GW ranged from 2.6 to 4.21 millimeter (mm), the GL ranged from 5.88 to 8.81 mm, the PH ranged from 65.08 to 148.3 centimeter (cm), the GNS ranged from 24 to 84.6, the SL ranged from 6.65 to 18.17 cm, and the SNS ranged from 15.83 to 27, respectively ( Table 1). The BLUP datasets of all traits showed normal distributions in the RIL lines, which suggested polygenic inheritance of these traits (Fig. 1A). Additionally, the TGW, GL, PH, GNS and SL showed high acrossenvironment broad-sense heritability of 0.54, 0.6, 0.91, 0.66 and 0.88, respectively (Table 1). Significant and positive correlations (P < 0.01) of the seven yield-related traits among all environments and the BLUP datasets were detected, which suggested that these traits were environmentally stable and mainly controlled by genetic factors (Table S3).

Correlation analyses among different traits
The BLUP datasets of each trait was employed to assess their correlations in the CM42×CM39 RIL population. TGW had significantly positive correlation with GW, GL, PH and SL, and significantly negative correlation with GNS and SNS (P < 0.001) (Fig. 1). GW was significantly and positively correlated with GL (P < 0.001), weakly and positively correlated with SL (P < 0.05), significantly and negatively correlated with GNS and SNS (P < 0.001), and not correlated with PH, respectively ( Fig. 1). GL had significantly positive correlation with PH and SL (P < 0.001), significantly negative correlation with GNS (P < 0.001), and weakly negative correlation with SNS (P < 0.05) (Fig. 1). Significantly positive correlations between PH and SL, GNS and SNS, and SL and SNS (P < 0.001), weakly positive correlations between PH and SNS (P < 0.05), significantly negative correlations between PH and GNS (P < 0.001), and no correlations between GNS and SL were detected, respectively (Fig. 1). Grain weight per spike (GWS) is comprised by TGW and GNS in wheat.  Thus, we further analyzed the correlation between the seven yield-related traits and the GWS. The results showed that GWS was significantly positive and positively correlated with TGW, GW, GL, GNS, SNS and SL (P < 0.05), and no correlated with PH (Table S4).

QTL detection
Phenotypic data of the seven yield-related traits in each environment and the BLUP datasets were used for QTL detection, in which the BLUP datasets were treated as an additional environment. A total of 30 QTLs were identified in multi-environments and located on all chromosomes excepting 3B, 3D, 4B, 4D, 5D and 6B (Table 2). For TGW, two QTLs were detected on chromosomes 6A. QTgw.cib-6A.1 was detected in two environments and the BLUP datasets, explaining 9.89-16.38% of the phenotypic variance. QTgw.cib-6A.2 was a major QTL detected in four environments and the BLUP datasets and explained 15.31-23.75% of the phenotypic variance. Alleles of CM42 for the two QTLs contributed to higher TGW (Table 2).
Among the six QTLs for GL, two major QTL QGl.cib-3A and QGl.cib-6A were identified in five environments and the BLUP datasets, explaining 6.55-11.86% and 5.96-13.11% of the GL variation, respectively. The positive additive effects of the two QTLs on GL were contributed by CM42. The rest four minor QTLs were identified in two or three environments on chromosome 5A, 6D and 7D, explaining 5.17-11.34% of the GL variation. The favorable alleles of QGl.cib-5A.1, QGl.cib-5A.2, and QGl. cib-7D were derived from CM42, and that of QGl.cib-6D was derived from CM39 (Table 2).
Among the six QTLs for PH, QPh.cib-2D on chromosome 2D was a stable QTL and detected in five environments and the BLUP datasets, explaining 4.54-9.38% of the PH variation. The allele of CM39 contributed to higher PH. The rest five minor QTLs on chromosomes 1A, 4A, 5A, 5B and 6A were detected in two or three environments, explaining 3.8-11.37% of the PH variation. The positive alleles of QPh.cib-1A and QPh.cib-5B were from CM39, and that of QPh.cib-4A, QPh.cib-5A and QPh.cib-6A were from CM42 ( Table 2).
Two minor QTLs for GNS on chromosomes 2D and 6A were detected in two environments and the BLUP datasets and explained 4.97-6.46% and 6.56-7.73% of the GNS variation, respectively. Alleles from CM42 and CM39 at QGns.cib-2D and QGns.cib-6A, respectively, contributed to positive effects on GNS (Table 2).
For SL, four QTLs were detected on chromosomes 2D, 5A, 5B and 6A. A major QTL QSl.cib-2D was detected in eight environments and the BLUP datasets, explaining 6.18-14.89% of the SL variation. QSl.cib-5B was a stable QTL and detected in three environments and the BLUP datasets, explaining 3.79-5.96% of the SL variation. Alleles of CM39 for the two QTLs contributed to increase of SL. Two minor QTLs QSl.cib-5A and QSl.cib-6A were detected in two or three environments, explaining 3.47-7.8% and 5.63-5.9% of the SL variation, respectively. The positive alleles of the two QTLs were contributed by CM42 (Table 2).

QTL analysis and comparison with previous studies
Wheat yield-related traits are significantly associated with yield and typically show higher heritability than the yield itself, and thus, mining the genes or QTLs related to yield-related traits will be help for elucidating the genetic basis of wheat yield and facilitating the genetic improvement of varieties with high yield [5][6][7]. In the present study, a RIL population derived from two elite winter wheat varieties were used to dissect the genetic basis of variation for seven yield-related traits, including TGW, GNS, GW, GL, PH, SL and SNS. A total of 30 QTLs were identified in multiple environments, explaining 2.34-23.75% of the phenotypic variance (Table 2). Fourteen QTLs were identified for grain size and weight, including two for TGW, six for GW and six for GL. Among them, QTgw.cib-6A. 1 and QGl.cib-6A were co-located on chromosome arm 6AS, which was near to QTkw-6A. 1 and QTgw.cau-6A.4 [1,43]. QTgw.cib-6A.2 was located on chromosome arm 6AL and near to  (Table 3). PH and SL are important traits related to plant architecture and yield potential in wheat [12,56]. In the present study, six and four QTLs for PH and SL were identified, respectively. Among them, QPh.cib-2D and QSl.cib-2D were co-located in the same interval on chromosome arm 2DS, which was overlapped with the dwarfing gene Rht8 [31,51]. QPh.cib-4A and QPh.cib-5A were located near to two loci for PH reported by Luján Basile et al [52]. QPh.cib-6A on chromosome 6A and CM39 indicate the lines with the alleles from CM42 and CM39, respectively; *, ** and *** represent significance at P < 0.05, P < 0.01, and P < 0.001, respectively; ns represents non-significance was overlapped with the dwarfing gene Rht18 [53]. QSl. cib-5A on chromosome 5A was located near to QSL5A.3 detected by Liu et al [54]. For the rest four QTLs QPh. cib-1A, QPh.cib-5B, QSl.cib-5B and QSl.cib-6A, no stable QTL for PH and SL reported previously was overlapped with them, indicating they are likely novel (Table 3).

QTL cluster on chromosomes 2D and 6A
Numerous co-located QTLs associated with multiple traits have been reported in the previous studies [2,5,24,57,58], which are beneficial to improve breeding efficiency for multiple elite traits, and thus is favorable for pyramiding breeding. In the present study, three QTLs QSl.cib-2D, QPh.cib-2D and QGns.cib-2D were co-located in the interval of 8.4-29.35 Mb on chromosome arm 2DS ( Table 2). The allele of CM42 at the locus decreases SL and PH while increasing GNS. Additionally, the locus was overlapped with the dwarfing gene Rht8, which has been reported to associated with QTLs for PH, SL, SNS, GNS, spikelet compactness, TGW, and grain yield [12,51,[59][60][61]. Interestingly, no QTL  [55] for grain size and weight detected in the present study was overlapped with the locus, indicating it had no effect on grain size and weight. Given CM42 was bred by utilizing synthetic wheat germplasm [62], further studies, such as fine-mapping and map-based cloning are needed to future reveal the relationship between the locus and Rht8. However, the results in this study showed that the locus could be utilized in optimization PH with no penalty for grain size and weight in MAS. Two QTL clusters were detected on chromosome 6A in the present study. One comprised two QTLs, QTgw.cib-6A.1 and QGl.cib-6A, was located on chromosome arm 6AS (Table 2, Fig. 2). The other one comprised five QTLs, QTgw.cib-6A.2, QGw.cib-6A, QGns.cib-6A, QPh.cib-6A, and QSl.cib-6A, was located on chromosome arm 6AL ( Table 2, Fig. 2). The QTL cluster on chromosome 6AL was overlapped with the haplotype block encompassing TaGW2 and additional 2167 genes which was located between 187 Mb and 455 Mb on chromosome 6A and defined by Brinton et al [63]. Therefore, fine-mapping and map-based cloning is needed to dissect the relationships between TaGW2 and the QTL cluster on chromosome 6AL in the future study. For the QTL cluster on chromosome 6AS, which was located between 73.08 Mb and 82.67 Mb and far apart the haplotype block of TaGW2 [63], indicating that they are different loci for grain weight.

Additive effects of three major QTLs on TGW and GNS
Due to there is a trade-off between TGW and GNS, increasing one of them may not contribute to an increase in grain yield of wheat. We further analyzed the additive effects of three major QTLs, QPh/Sl.cib-2D, QGl.cib-3A and QTgw.cib-6A.2, on the TGW and GNS. As showed in the Table 4, lines possessing the allele from CM42 at the three loci had relatively higher TGW and GNS, which might partly explain the high yield of CM42. Additionally, lines possessing the alleles from CM42 at QPh/ Sl.cib-2D and QTgw.cib-6A.2 and the allele from CM39 at QGl.cib-3A also had relatively higher TGW and GNS. However, for the other combination schemes, either the higher TGW but lower GNS, or higher GNS but lower TGW, or both lower TGW and GNS were harvested. Overall, the QTLs and KASP markers in this study will be useful for elucidating the genetic architecture of grain yield and developing new wheat varieties with high and stable yield in wheat.
aa, bb and cc represent the allele from CM39 at QPh/ Sl.cib-2D, QGl.cib-3A and QTgw.cib-6A.2, respectively; AA, BB and CC represent the allele from CM42 at QPh/ Sl.cib-2D, QGl.cib-3A and QTgw.cib-6A.2, respectively; Lines represent the number of different haplotypes; * and ** represent significance at P < 0.05 and P < 0.01, respectively; The superscript letter indicates significant difference among groups  Fig. 2). In the interval of QTgw.cib-6A.1/ QGl.cib-6A and QGl.cib-3A, there were 81 and 85 predicted genes in the CS genome, respectively (Table S6,  S7). Expression pattern analyses showed that 45 and 57 genes in the interval of QTgw.cib-6A.1/QGl.cib-6A and Gl.cib-3A expressed in various tissue, respectively (Fig. 4) [65,66]. Among them, several were abundantly expressed in grain, indicating they are likely associated with grain growth and development (Fig. 4). For example, TraesC-S6A02G107800 is an ortholog of the rice RGG2 and encodes a guanine nucleotide-binding protein subunit gamma 2 (Table S6). Miao et al previously reported that RGG2 played a negative role in plant growth and yield production and that manipulation of RGG2 can increase the plant biomass, grain weight, length and yield in rice [67]. TraesCS6A02G112400 and TraesCS3A02G424000 encode polyubiquitin and small ubiquitin-related modifier, respectively (Table S6, S7). TraesCS3A02G421900 encodes a 26S proteasome regulatory subunit (Table S7), which participates in the ubiquitin/26S proteasome pathway and mediate the degradation of the complex of ubiquitin receptor and poly-ubiquitinated protein [68,69]. Previous studies revealed that the ubiquitin pathway play an important role in regulation grain size and weight in rice [70,71]. These results indicated that the four genes may be closely related to grain size and weight in wheat and useful for fine mapping and cloning of QTgw.cib-6A.1/QGl.cib-6A and QGl.cib-3A in our following work.

Conclusion
In

Plant materials and field trials
A RIL population (F 10 ) comprising 193 lines derived from a cross CM42 and CM39 were used for QTL detection in the present study. CM42 is the first wheat elite variety in the world bred by using synthetic hexaploid wheat (Triticum turgidum×Aegilops tauschii) germplasm, and showed high yield potential in Sichuan and the Yangzi River region [62], while CM39 is an elite winter wheat variety with different genetic background to that of CM42. During four growing seasons of wheat

Phenotyping and statistical analysis
At maturity, ten representative plants from middle row of each line were randomly selected to investigate agronomic traits including TGW, GL, GW, GNS, PH, SL, SNS and GWS. SL was measured as the length from the base of the rachis to the tip of the terminal spikelet, excluding the awns. SNS was determined by counting the number of spikelets in main spikes; PH was measured from the soil surface to the tip of the spike, excluding the awns. Subsequently, the main spike of all selected plants were harvested and manually threshed for evaluating GNS, TGW, GW, GL and GWS using SC-G software (Wseen Co., Ltd, Hangzhou, China). PH and SL were evaluated in eight environments, and the rest traits were evaluated in six environments. Basic phenotypic statistical analyses, frequency distribution, correlation analyses and student's t tests were performed with SPSS version 20.0 (Chicago, IL, USA). The phenotype distribution graph was drawn using the plugin "CorrPlot" in TBtools [72]. The relationships among measured traits were visualized using the R package "qgraph". The BLUP data across evaluated environments was calculated using the "lmer" function implemented in R package "lme4". ANOVA was performed over all trials which indicated statistically significant main effects for genotypes (G), environments (E), G × E interactions for all measured traits using the SAS software (SAS Institute Inc., North Carolina, USA). The broad sense heritability (H 2 ) was estimated based on the following equation: H 2 = σ 2 g / σ 2 g + σ 2 ge /n + σ 2 e /nr , whereσ 2 g is the variance of genotypes, σ 2 ge is the variance of genotype by environmental effect, σ 2 e is the residual variance, n is the number of environments and r is the number of replicates [73].

Linkage map construction and QTL detection
A whole-genome genetic map constructed previously was adopted for QTL mapping [42]. The genetic map was constructed using the CM42×CM39 RIL population with SLAF markers. A total of 4996 Bin SLAFs were distributed in 21 linkage groups and covered a total genetic distance of 2,859.94 cM with an average interval of 0.57 cM between adjacent Bin marker (Table S1, S2) [42].
QTL analysis was conducted using the inclusive composite interval mapping (ICIM) function of IciMapping 4.1 (https:// www. isbre eding. net) with the minimal LOD score was set at 2.5. The missing phenotype was deleted in QTL analysis. QTL was named according to the provision of Genetic Nomenclature (http:// wheat. pw. usda. gov/ ggpag es/ wgc/ 98/ Intro. htm), where 'CIB' represents Chengdu Institute of Biology. QTLs consistently identified in at least three environments and in combined analysis with ≥10% of phenotypic variation explained were considered as major QTLs.

Development of Kompetitive Allele-Specific PCR Markers
On the basis of the preliminary QTL mapping results, the flanking markers of major QTL were blasted against the CS reference genome sequence (RefSeq v1.0; https:// wheat-urgi. versa illes. inra. fr/) to gain their physical positions [74]. The SNPs within the physical interval of major QTLs were used for developing KASP markers tight linked with them. The KASP marker primers were designed using the PrimerServer tool in Triticeae Multi-omics Center (http:// 202. 194. 139. 32/) [75]. Standard FAM and HEX adapters were added to the allele-specific forward primers at the 5′ ends. The KASP assays were run in a Bio-Rad CFX96 real-time PCR system in 10μL reaction volumes with the following PCR cycling parameters: hot start enzyme activation at 94 °C for 15 min; a touchdown of 10 cycles (94 °C for 20 s, and touchdown starting at 61 °C and decreasing by 0.6 °C per 1-min cycle); then 26 cycles of regular PCR (94 °C for 20 s, 55 °C for 60 s, and rest at 37 °C for 1 min). If the clustering was not significant, further cycling was performed at 94 °C for 20 s and 55 °C for 60 s (3-10 cycles per step)

Prediction of candidate gene
Genes between the physical intervals of major QTLs were extracted from IWGSC RefSeq v1.1 annotation for CS [74]. The annotations and functions of a given gene were analyzed using UniProt (https:// www. unipr ot. org/). The expression pattern analysis was performed by using Wheat Expression Browser (http:// www. wheat-expre ssion. com/), and the circle graph of expression values was drawn using TBtools [72]. The orthologous gene analysis between wheat and rice was conducted using the Triticeae-Gene Tribe (http:// wheat. cau. edu. cn/ TGT/) [76].