The recombinant inbred line (RIL) population of 163 F₆ lines was used for QTL analysis of kernel traits in this study, derived from the cross Avocet × Chilero using the single seed descent approach (Basnet et al., 2014). Chilero had significantly higher values (p < 0.05) for all investigated kernel traits than those of Avocet. The International Maize and Wheat Improvement Center (CIMMYT) developed the RIL population.

A natural population with 114 cultivars was utilized for validation of the KASP markers, including 53 wheat accessions collected within the country and 61 cultivars from other countries, such as those in Europe, USA, Mexico, and Australia. Materials were provided by the wheat germplasm innovation and molecular breeding project of Henan University of Science and Technology, China.

For phenotyping, the RIL population and parents were grown under four environments at three experimental stations in three seasons: (1) a test field at the farm of Henan University of Science and Technology (34°C60N″, 112°C42E″), during the autumn–winter cycle of 2018–2019 (2018XN) and 2020–2021 (2020XN) cropping seasons; (2) an open field at Luoning county (34°C42″N, 111°C67″E) in the 2019–2020 (2019LN) cropping season; and (3) a test field at Mengjin county (34°C83″N, 112°C58″E) during the 2020–2021 (2020MJ) cropping season. The natural population of 114 varieties or lines was planted in the three experimental stations at Mengjin county and the farm of Henan University of Science and Technology during 2019–2021.

Each trial of both populations was arranged following a randomized block design with three replicates; the lines of the RIL population and the natural population were grown in a 6.0 m² plot at each location, with 10 rows, 20 cm apart and 3.0 m in length for each plot. Field management followed the local agronomic practices.

In each experiment, plants were chosen to be harvested after they were completely matured, in order to avoid other factors that may affect the phenotypic analyses. Meanwhile, seeds were dried before analysis. Data parameters of kernel traits were evaluated when all the kernels had approximately 11% moisture content.

The phenotypic values of two populations were determined using the same method. Three kernel traits (TKW, KL, and KW) were measured by using Wanshen SC-A automatic testing equipment, which was developed by Wanshen Science and Technology Ltd. (Hangzhou, Zhejiang, China; www.wseen.com). At least 300 kernels from each line were measured with three replicates, and the average of the three replicates was taken as the evaluation result of each line.

For each phenotypic trait, SPSS version 22.0 software (SPSS, Chicago, USA) was used to conduct statistical analysis of phenotypic data, including the means, standard deviation (SD), range, kurtosis, skewness, and coefficient of variation. The QTL IciMapping v4.1 software was used to compute the broad-sense heritability (H²) and to calculate the best linear unbiased estimate (BLUE) of each kernel trait. The OriginPro 22b software was used to draw the histograms and correlation plot, and the linkage map was drawn using Mapchart.

A total of 23,526 DArTSeq markers were genotyped for both parents and the RIL population, and the QTL IciMapping v4.1 software was used to construct the genetic linkage map and identify significant QTLs (http://www.isbreeding.net) (Zeng et al., 2020). The Kosambi mapping function was used to calculate centiMorgan units (cM), and the inclusive composite interval mapping (ICIM) method was performed for QTL analysis. For all significant QTLs, the critical LOD values were set at 3.0 to increase the reliability and accuracy of QTL detection, and the walking speed parameter of each step for the genome-wide scan was set at 1.0 cM; the significance thresholds were calculated using 1,000 permutations, with genome-wide error rates of 0.10 and a type I error of 0.01. The naming of QTL followed the rule “QTL + trait + research department + chromosome”. The QTLs detected in two or more environments are considered as stable QTLs (Ruan et al., 2021).

In this study, we performed the candidate gene analysis for a highly significant and stable region on wheat chromosome 1AS (http://plants.ensembl.org/index.html). The high-confidence genes were extracted from the IWGSC reference genome and identified using IWGSC RefSeq v1.1 (https://urgi.versailles.inra.fr/) annotation for the identification of likely candidates.

Based on the mapping results, the sequences flanking the QTL TaTKW-1A were used for designing KASP primers (composed of the two forward primers and the reverse primer) (PolyMarker, http://polymarker.tgac.ac.uk/). The primers were synthesized by Sangon Biotech (Shanghai) Co., Ltd. (China).

All KASP reactions were performed in a 4-μl reaction volume, which included 2 μl of diluted DNA, 2 μl of KASP master mix, and 0.045 μl of primer mix. A total of 114 wheat varieties were genotyped on an CFX 384 Real-Time System (BIO-RAD). The fluorescence signals of each reaction well were collected and genotyping was performed using the BioRad CFX Manager Software.

In the four field trials conducted, the means, standard deviation (SD), range, kurtosis, skewness, and coefficient of variation for each of the phenotypes were calculated in the RIL population. The parental genotype Chilero of the mapping population consistently had significantly higher mean values (p < 0.05) for all investigated kernel traits than those of Avocet ( Table 1 ). According to the phenotypic distribution, TKW was larger than KL and KW on the range of variation, and the scores of skewness and kurtosis were mostly less than 1.0 for all kernel traits in the four field trials, indicating that they were quantitative traits controlled by multiple genes. All kernel traits had broad-sense heritability higher than 90% ( Table 1 ).

Continuous variations and strong transgressive segregations have been shown for all three traits in the RIL population, suggesting that favorable alleles of these traits are distributed in both parents, and indicating segregation patterns of quantitative traits ( Figures 1 , 2 ). Correlations among kernel traits are significant ( Figure 3 ).

In this study, we obtained the results using the ICIM method; a total of 48 QTLs for TKW, KL, and KW were detected and mapped on chromosomes 1A, 1B, 1D, 2B, 2D, 3A, 3B, 3D, 4A, 4B, 5A, 5D, 6A, 6B, 6D, 7A, 7B, and 7D (Tables S1, S2, S3; Figures S1, S2, S3), including 12 QTLs for TKW, 18 QTLs for KL, and 18 QTLs for KW, individually explaining 3.00%–33.85% of the phenotypic variance. Among these QTLs, 9 QTLs for TKW, 12 QTLs for KL, and 13 QTLs for KW explained >10% of the phenotypic variance and were considered the major QTLs.

The sequences of the flanking markers of the QTLs were employed to perform BLASTN against the Chinese Spring reference genome sequence v1.1. According to the physical location of the QTLs, we identified the stable QTL clusters in the Avocet/Chilero RIL population; the detailed information is described in Table 2 and Figure 4 .

The results showed that there were eight stable QTL clusters in the population, and they were renamed TaTKW-1A, TaKW-3A, TaTKW-3B, TaKW-3B, TaTKW-4A, TaKW-4B, TaTKW-5A, and TaKL-6A. Among the eight stable QTL clusters, four QTL clusters have the favorable allele from Avocet, TaTKW-1A, TaKW-3A, TaTKW-4A, and TaKL-6A, and four QTL clusters have the favorable allele from Chilero, TaTKW-3B, TaKW-3B, TaKW-4B, and TaTKW-5A. TaTKW-1A was closely linked to DArT markers SNP1158610, SNP1090977, 1010408, SNP100461256, 3950546, and 1213099 in the physical interval of 14.56–49.30 Mb, explaining 10.31%–33.85% of the phenotypic variance. TaKW-3A was tightly linked to marker intervals SNP1093344–100008048 with a physical interval of 69.50–69.51 Mb, accounting for 7.76%–20.42% of the phenotypic variance.

To clarify the physical position of TaTKW-1A, the DArT marker sequence was subjected to alignment with the whole-genome database of Chinese Spring (https://wheat-urgi.versailles.inra.fr/) by using the BLAST tool. Sequence comparison revealed that TaTKW-1A was in a physical interval from 14557761 to 49301348 bp on chromosome 1A ( Figure 5 ). A total of 347 high-confidence genes with a physical length of 34.74 Mb were identified in the DarT marker interval 3950546–1213099 (https://www.wheatgmap.org/).

For the effective utilization of the major QTL in plant breeding, KASP markers closely linked to TaTKW-1A were developed and used to genotype 114 lines ( Table 3 , Figure 6 ).

Of the 114 accessions, there were 26 GG (22.8%) genotypes and 88 AA (77.2%) genotypes ( Figure 6 ), and TKW was significantly different (p < 0.01) between the two genotypes; in addition, AA (Avocet) genotypes were higher than GG (Chilero) genotypes. Furthermore, 15 wheat varieties have genotype GG and 38 have genotype AA in the 53 domestic wheat accessions, and 11 have genotype GG and 50 have genotype AA in the 61 foreign varieties (Figure S4).

Wheat has a very huge and complex genome; QTL mapping can provide important information regarding the molecular basis of determination of kernel-related traits. In past decades, more than 400 QTLs for TKW and approximately 200 QTLs for KL and KW have been reported across all 21 chromosomes, and some stable and robust QTLs were detected (Saini et al., 2022). As a whole, these QTLs were mostly distributed across the A and B genomes as compared to the D genome (Yang et al., 2020). A similar trend was observed in this study, with more QTLs on the A (24) and B (14) genomes than on the D (10) genome. Although many QTLs associated with kernel traits have been identified, its application is rarely reported in molecular marker-assisted breeding (MAS), due to the fact that many QTL locations were based on genetic distances rather than the physical distances.

TKW was a complex quantitative trait that was affected by polygenes. Studies indicated that TKW increased gradually when KL and KW increased, and TKW had a significant positive correlation with KL and KW (Gao et al., 2015; Cui et al., 2016; Chai et al., 2022). In this study, the significant correlation is found between TKW and KL, between TKW and KW, and between KL and KW, with the Pearson correlation coefficients of 0.51, 0.90, and 0.20, respectively, which was consistent with the conclusions of other studies (Liu et al., 2014; Michel et al., 2019; Qu et al., 2022).

TKW was a complex polygenic trait with high broad-sense heritability and was less affected by the environment (Cuthbert et al., 2008; Mcintyre et al., 2010; Gao et al., 2015), and QTLs for TKW have been reported on all 21 wheat chromosomes (Huang et al., 2006; Li et al., 2007; Sun et al., 2008; Ramya et al., 2010; Zhang et al., 2014; Shukla et al., 2015; Yu et al., 2018; Chen et al., 2020). In this study, we detected four QTL clusters for TKW, designated TaTKW-1A, TaTKW-3B, TaTKW-4A, and TaTKW-5A. Then, based on the sequence information of the markers flanking these QTLs, we found some genes within each of these QTL intervals by using the BLAST tool.

In earlier studies, many QTLs for TKW were reported on 1A (Pinto et al., 2010; Ramya et al., 2010; Williams and Sorrells, 2014; Wu et al., 2015; Cui et al., 2016; Assanga et al., 2017), 3B (Russo et al., 2014; Xu et al., 2014; Shirdelmoghanloo et al., 2016; Li et al., 2018), 4A (Zou et al., 2017; Li et al., 2020), and 5A (Liu et al., 2014; Fowler et al., 2016; Kumar et al., 2016; Assanga et al., 2017; Brinton et al., 2017). Among the four loci, TaTKW-1A colocalized with QTKW.haust-1A.1, QKW.haust-1A.2, QTKW.haust-1A.2, and QKL.haust-1A.2, with a corresponding physical position of 14557761–49301348 bp on the short arm of chromosome 1A. It overlapped with few published QTLs, such as QGw.ccsu-1A.1 (physical position: 27.27 Mb), QTkw.ncl-1A.1 (physical position: 27.27 Mb), and qTgw.nwipb-1AS (physical interval: 20.02–25.31 Mb), which was detected by Kumar et al. (2006); Ramya et al. (2010), and Liu et al. (2020a), respectively. However, these QTLs were only detected in one environment, and further studies of the locus have not been reported in the literature. TaTKW-4A colocalized with QTKW.haust-4A and QKL.haust-4A.2. The physical interval (584.41–606.37 Mb) of the loci was not identical to that previously identified (Maphosa et al., 2014; Gao et al., 2015; Chen et al., 2020; Li et al., 2020); these loci may be novel loci for TKW. Another colocated QTL, TaTKW-5A (QTKW.haust-5A.1, QTKW.haust-5A.2, and QKW.haust-5A) at 436.63–481.93 Mb, does not coincide with the physical position of other reported QTLs (Wang et al., 2009; Yuan et al., 2017; Guan et al., 2018). TaTKW-3B colocalized with QKW.haust-3B.1 and QTKW.haust-3B.1, and it was localized to the physical interval 7.29–31.66 Mb. We have undertaken a detailed analysis of the loci identified on chromosome 3B in previous studies (Huang et al., 2004; Maphosa et al., 2014; Echeverry-Solarte et al., 2015; Shi et al., 2017; Duan et al., 2020); TaTKW-3B was different from those QTLs reported on 3B, and was likely a new locus.

With respect to KL and KW, many QTLs for KL and KW were previously reported on chromosomes 3A, 3B, 4B, and 6A (Breseghello and Sorrells, 2006; Gegas et al., 2010; Cui et al., 2016; Cheng et al., 2017; Li et al., 2018; Zhai et al., 2018). In our study, we detected one QTL cluster for KL on chromosome 6A (TaKL-6A, physical interval: 61.21–65.66 Mb) and three QTLs for KW on chromosomes 3A (TaKW-3A, physical interval: 69.5–69.51 Mb), 3B (TaKW-3B, physical interval: 490.89–614.57 Mb), and 4B (TaKW-4B, physical interval: 629.7–630.13 Mb). To confirm whether we identified four loci, a comparative analysis of the physical positions of previously reported QTLs with those identified in this study was conducted. According to the physical intervals of TaKL-6A, TaKW-3A, TaKW-3B, and TaKW-4B, we speculated that the four loci were new QTLs (Cui et al., 2011; Cui et al., 2014; Li et al., 2018; Su et al., 2018; Chen et al., 2020; Duan et al., 2020; Xin et al., 2020).

In recent years, with the rapid development of sequencing technology and bioinformatics, a fully annotated reference genome of Chinese Spring was released (IWGSC RefSeq v1.1, https://urgi.versailles.inra.fr/blast_iwgsc/blast.php), providing a better approach in searching for candidate genes. Meanwhile, due to the co-linearity with grasses and the conservation of gene function among different species, many functional markers in wheat have been developed for many cloned genes of kernel traits.

In TaTKW-1A, blasting results showed a physical interval of 14.56–49.30 Mb, and a total of 347 high-confidence genes were found ( Figure 5 ). Among these genes ( Figure 7 ), TraesCS1A02G045300 and TraesCS1A02G058400 were the most promising candidate genes associated with kernel weight, and their orthologs were Os05g0115800 and Os05g0121600 in rice, respectively. Os05g0115800 was involved in the mitogen-activated protein kinase signaling pathway and was a mitogen-activated protein kinase phosphatase, affecting grain yield by regulating the grain number and grain size (Jiang et al., 2019). Os05g0121600 was involved in the regulation of transcription, flower development, seed development, and endosperm development, and acted as a negative regulator in starch synthesis (Seetharam et al., 2021). In this study, we speculated candidate genes for TKW on chromosomes 4A and 5A. The physical interval of TaTKW-4A was 584.41–606.37 Mb, and 406 annotated genes were presumed. TraesCS4A02G293900, TraesCS4A02G294000, and TraesCS4A02G303500 were the candidate genes for TKW, and the orthologs were Os03g0669100, At4g34460, and At3g21510, respectively. Among these genes, Os03g0669100 and At4g34460 encoded a regulator of G-protein signaling (RGB1 and AGB1) (Oliveira et al., 2022), and At3g21510 (AHPs) and its encoded protein are related to the regulation of endosperm growth (Tran et al., 2021). RGB1, AGB1, and AHPs were involved in the regulation of grain traits (Li et al., 2019; Li et al., 2021). In TaTKW-5A, the physical interval was 436.63–481.93 Mb, and there were 512 high-confidence genes. Os09g0448500 of the ortholog of TraesCS5A02G233400 was a transcriptional regulator in rice and was associated with kernel traits (Li et al., 2019; Zhou and Xue, 2020). Consequently, TraesCS4A02G293900, TraesCS4A02G294000, TraesCS4A02G303500, and TraesCS5A02G233400 were the candidate genes on 4A and 5A chromosomes in wheat ( Figure 7 ). In TaTKW-3B, the physical interval was 7.29–31.66 Mb; a total of 418 annotated genes were found in the physical intervals, but we did not find orthologs related to TKW.

Previous studies have shown that the growth of maternal tissues is able to control seed size through several signaling pathways, including the ubiquitin–proteasome pathway (Huang et al., 2017; Xie et al., 2018), G-protein signaling (Liu et al., 2018; Sun et al., 2018), mitogen-activated protein kinase signaling (Guo et al., 2018; Xu et al., 2018), phytohormones (Xu et al., 2015; Zhou et al., 2017), and transcriptional regulators (Wang et al., 2015; Segami et al., 2017). In our study, we found TaTKW-1A, which has two major candidate genes, TraesCS1A02G045300 and TraesCS1A02G058400, with grain-related traits ( Figure 7 ). We identified them in the transcriptome of wheat grain through the website https://www.ebi.ac.uk/gxa/home (Gillies et al., 2012; Li et al., 2013; Takafuji et al., 2021; Yu et al., 2021) (Figure S5). Results showed that both genes were expressed during grain development, although the expression profiles of these two genes clearly differ during grain development; both of them were expressed in the pericarp, endosperm, and seed coat (Li et al., 2013; Yu et al., 2021). Specifically, TraesCS1A02G045300 is important because it was expressed consistently from anthesis to maturity (Pfeifer et al., 2014; Pearce et al., 2015; Yu et al., 2021).

With the rapid development of marker-assisted selection in wheat breeding, the molecular marker technology has received increasing attention in recent years in crops (Song et al., 2022), and numerous markers have been developed (Kang et al., 2020; Rambla et al., 2022; Shin et al., 2022). A CAPS marker, TaTPP-6AL1-CAPS, was developed to differentiate TaTPP-6AL1a and TaTPP-6AL1b, which was associated with TKW (Zhang et al., 2017). A KASP marker of the candidate gene TaFT-D1, which was associated with TKW and KW, was developed and verified in a natural population (Liu et al., 2020b). Ten SSR markers for grain weight were developed and tested in 60 genotypes; all SSR primers had a high polymorphism (Sallam et al., 2019). In particular, numerous KASP markers for kernel traits have been developed in the last 2 years: Kasp_5B_Tgw for QTgw.caas-5B was developed and validated in wheat (Zhao et al., 2021), a KASP functional marker of TaTAP46-5A associated with kernel weight in wheat was developed and identified (Zhang et al., 2021), the KASP markers for QTKW.caas-5DL (Song et al., 2022), and the KASP markers for QGl.cib-4A (Li et al., 2022). These KASP markers provide a robust tool for genetic mapping and molecular breeding in crops. In this study, in order to use the advantage haplotypes, we developed the KASP markers of TaTKW-1A, which have been validated in the natural population. The results showed that the KASP markers could be used in wheat.