Analysis of Genetic Regions Related to Field Grain Number per Spike From Chinese Wheat Founder Parent Linfen 5064

Wheat founder parents have been important in the development of new wheat cultivars. Understanding the effects of specific genome regions on yield-related traits in founder variety derivatives can enable more efficient use of these genetic resources through molecular breeding. In this study, the genetic regions related to field grain number per spike (GNS) from the founder parent Linfen 5064 were analyzed using a doubled haploid (DH) population developed from a cross between Linfen 5064 and Nongda 3338. Quantitative trait loci (QTL) for five spike-related traits over nine experimental locations/years were identified, namely, total spikelet number per spike (TSS), base sterile spikelet number per spike (BSSS), top sterile spikelet number per spike (TSSS), fertile spikelet number per spike (FSS), and GNS. A total of 13 stable QTL explaining 3.91–19.51% of the phenotypic variation were found. The effect of six of these QTL, Qtss.saw-2B.1, Qtss.saw-2B.2, Qtss.saw-3B, Qfss.saw-2B.2, Qbsss.saw-5A.1, and Qgns.saw-1A, were verified by another DH population (Linfen 5064/Jinmai 47), which showed extreme significance (P < 0.05) in more than three environments. No homologs of reported grain number-related from grass species were found in the physical regions of Qtss.saw-2B.1 and Qtss.saw-3B, that indicating both of them are novel QTL, or possess novel-related genes. The positive alleles of Qtss.saw-2B.2 from Linfen 5064 have the larger effect on TSS (3.30%, 0.62) and have 66.89% in Chinese cultivars under long-term artificial selection. This study revealed three key regions for GNS in Linfen 5064 and provides insights into molecular marker-assisted breeding.


INTRODUCTION
Founder parents are not only successful cultivars that are cultivated in large areas but are also used extensively as parents in breeding programs. These valuable genetic resources are crucial to Chinese wheat breeding programs (Zhuang, 2003). Analyzing the genetic diversity of founder parents and the genetic basis of their widespread success can provide a foundation for more efficient use of these germplasm resources.
A Chinese wheat founder parent named Linfen 5064 is the pedigree of more than 80 high-quality strong gluten cultivars in China. Linfen 5064 has the strong-gluten trait, a high grain number per spike (GNS), and excellent agronomic traits (Qiao et al., 2018). Linfen 5064 and cultivars derived from it not only have high yields but have also been used as the main parents for improving wheat quality in Chinese breeding programs. The use of Linfen 5064 as the founder parent addressed three difficult points in the breeding for strong-gluten wheat (Qiao et al., 2018). The first difficultly is that quality is negatively correlated with GNS and thousand kernel weight (TKW). Chinese wheat cultivars with premium grain quality, such as Xinong 20, Fengdecun 5, Shiluan 02-1 and Jimai 20, usually have lower GNS and lower yields. The GNS of Linfen 5064 and cultivars and lines derived from it have higher yields than other high-quality cultivars. The second difficult point is that dwarfism is associated with late maturity. Linfen 5064 does not show this association as it matures early and is a semi-dwarf height of about 75 cm. Finally, Linfen 5064 overcomes the need to have the glutenin subunit combination 5 + 10 for good quality, since it lacks these subunits yet still has good quality. Therefore, the utilization of valuable traits of Linfen 5064, and the successful future breeding program of Wheat, it is essential to explore and analyze their genetic base.
In most wheat cultivars, a spike usually generates more than 10-20 spikelets, and each spikelet can differentiate into 9-10 florets (Cui et al., 2008). The differentiation of bract and floret primordia determines the number of spikelets and initial florets. During floret development, 60-80% of the initial florets either abort or otherwise lose fertility (Guo et al., 2015). The number of surviving florets which can eventually develop into grains determines the number of grains per spike . GNS shows high heritability (Isham et al., 2021). Increasing GNS is an important way to increase grain yield. GNS can be divided into total spikelet number per spike (TSS), fertile spikelet number per spike (FSS), base sterile spikelet number per spike (BSSS), top sterile spikelet number per spike (TSSS), and grains per spikelet. The heritability of TSS was higher (Isham et al., 2021), but the number of grains per spikelet and spikelet propagation ability were greatly affected by the environment. The map-based cloning of common wheat genes lags that of other crops because of wheat's large genome size. Consequently, most studies focus on the quantitative trait loci (QTL) level of analysis, especially genes/QTL that control yield traits.
Although many QTL/genes associated with GNS have been reported in wheat, the major and stable QTL identified under multiple environments are still limited. In addition, the biparents used for mapping were mostly accessions aim at certain traits rather than founder cultivars, the use of QTL identified need long-term backcross process which is time-consuming and low efficiency. We especially used founder parent and core cultivars in breeding as biparents for mapping, the loci obtained and markers developed are easily used in breeding, also provide evidence on utilization of the derivatives. Two doubled haploid (DH) populations (Linfen 5064 × Nongda 3338 and Linfen 5064 × Jinmai 47) were analyzed for five GNS-related traits over the nine experimental locations/years to (1) identify and validate major, stable QTL for GNS that can be used for molecular marker-assisted breeding and (2) identify genetic regions associated with GNS of Linfen 5064, elucidate the genetic mechanism of GNS in the founder parent, and discover favorable allele variations.

Plant Materials
A total of two DH populations were used, 192 lines from the cross Linfen 5064 × Nongda 3338 (LN) and 194 lines from the cross Linfen 5064 × Jinmai 47 (LJ). Linfen 5064 is a Chinese wheat founder parent with strong gluten, a high GNS, and an excellent array of other characteristics (Qiao et al., 2018). Nongda 3338, developed by China Agricultural University, is a "core parental" breeding line for the North China Winter Wheat Breeding Program with high general combining ability and the dwarfing genes Rht-B1b and Rht-D1b (Kabir et al., 2015). Jinmai 47 has the advantages of drought tolerance, stable yield, and a high utilization rate of water and fertilizer (Song et al., 2017). The phenotypic difference between the two cultivars and Linfen 5064 was significant and there was obvious trait separation in the population. LN was used for QTL analysis and LJ was used to validate the effects of putative QTL identified in LN.

Field Evaluation
The two DH populations were planted as a single replication in three locations in 2018-2019, 2019-2020, and 2020-2021 (19 YC,20YC,and 21 YC). The seed was sown in two 1.5 m rows per line spaced 0.3 m apart at 21 seeds per row. Field management practices were those commonly used in wheat production in the region.

Phenotypic Evaluation and Data Analysis
Ten days before harvest, data of five spike traits, TSS, BSSS, TSSS, FSS, and GNS, were collected by randomly choosing 10 plants in each line. FSS = TSS-BSSS-TSSS. The best linear unbiased prediction (BLUP) of target traits in different environments (Smith et al., 1998) and the broad-sense heritability (H 2 ) were obtained using SAS (SAS Institute, Cary, NC, USA; https://www. sas.com). The SPSS18.0 software (SPSS, Chicago, Illinois, USA; http://en.wikipedia.org/wiki/SPSS) was used to perform Student's t-test (p < 0.05) and correlation analysis of phenotype values in different environments.

Genetic Map Construction and Linkage Analysis
The two DH and parental lines were genotyped with a 15 K single-nucleotide polymorphism (SNP) panel developed based on 20 resequencing datasets, 1,520 genotyping datasets collected globally from multiple platforms, and publicly released resequencing and exon capture data. These datasets were developed and optimized using GenoBait technology to finally yield 14,868 mSNP regions for use in this study. The genetic map of LN was constructed using IciMapping 4.1 (Meng et al., 2015) and JoinMap 4.0. Markers were binned if the correlation coefficient between them was 1 using the BIN function in IciMapping 4.1 according to the method reported by Winfield et al. (2016). WinQTLCart version 2.5 (Wang et al., 2012) for composite interval mapping was used to detect QTL. The minimal logarithm of odds (LOD) score to accept the presence of a QTL was set at 2.5. QTL was considered major when more than 10% of the phenotypic variation was explained in at least one environment and it was detected in at least three environments, including the BLUP dataset. QTL either <1 cM apart or sharing common flanking markers were treated as a single locus.

Validation for the Major QTL Identified
Peak SNPs for stable QTL identified in the LN population were genotyped in the LJ population. The differences in spike-related traits between both groups in the LJ population were analyzed with a t-test in SAS V8.0.

Genes Identified in the Major QTL
Genes within the target region of major QTL were obtained using the genome browser (JBrowse) on the WheatOmics-bata website http://wheatomics.sdau.edu.cn/ (Ma et al., 2021). Functional annotation and enrichment analysis of genes in these regions were done using the gene ontology (GO) database and the R package cluster Profiler. Analysis of orthologs between wheat and rice used the Triticeae-Gene Tribe website (http://wheat.cau. edu.cn/TGT/). The expVIP public database (http://www.wheatexpression.com/) was used to search for the expression data of genes in 16 tissues and organs, perform log2 conversion processing, and analyze the expression patterns of genes.
The R software package LD heatmap of major QTL was used to draw the linkage disequilibrium heatmap according to the resequencing data in 145 landmark cultivars that were

Phenotypic Variation and Correlations of Five Traits in Nine Environments
Linfen 5064 had lower values for TSS and TSSS, and a higher value of GNS than Nongda 3338 ( Table 1). The spike traits of the DH population showed continuous variation, suggesting multigene genetic control. The estimated H 2 of five traits ranged from 0.78 to 0.92, indicating that these traits were significantly affected by genetic factors ( Table 1). The Pearson correlation coefficients among different environments were significant (P < 0.05, Supplementary Table S1). Better among-environment correlations were observed for TSS than for FSS, TSSS, BSSS, and GNS. Phenotypic correlations among spike traits were evaluated using the BLUP dataset ( Table 2). GNS significantly and positively correlated with FSS and TSS. GNS and FSS significantly and negatively correlated with BSSS and TSSS (p < 0.01, Table 2). The order of correlation coefficient with GNS were FSS (0.630) > TSSS (−0.437) > TSS (0.336) > BSSS (−0.162). These results showed that FSS and TSSS exerted great influence on GNS.

Linkage Map Construction
In total, 841 SNP markers were used for constructing the LN genetic map. The map had 21 linkage groups, a total length of 3045.86 cM, and an average interval distance of 3.62 cM. The D genome had the lowest marker coverage, especially for chromosomes 5D and 6D. The maps of the A, B, and D genomes had, respectively, lengths of 1324.20, 1322.53, and 399.14 cM and densities of 3.99, 3.28, and 3.77 cM/marker (Supplementary Table S2).

QTL for Spikelet Number per Spike
A total of 64 QTL for TSS, FSS, TSSS, and BSSS were detected on 18 chromosomes (Supplementary Table S3) with 13 stable QTL identified (Table 3). QTL were found on all chromosomes except 1D, 6D, and 7D (Supplementary Table S3

QTL for Grain Number per Spike
For GNS, 16 QTL were detected and these QTL explained 4.18-15.83% of the phenotypic variance (Supplementary Table S3). Four stable QTL, Qgns.saw-5B.2, Qgns.saw-7A.1, Qgns.saw-4D, and Qgns.saw-1A, explaining 4.47-11.16% of the phenotypic variance were identified in more than three environments and with BLUP values ( Table 3). The additive effect of Qgns.saw-7A.1 was from Linfen 5064 indicating that Linfen 5064 contributed the allele for increased GNS. No stable QTL clusters for GNS and spikelet number per spike were detected on the same chromosome, indicating that the QTL of GNS were most likely independent of spikelet number per spike and therefore have great potential in wheat breeding.

QTL Validation
To further validate the stable QTL, the peak SNPs for each were used to evaluate their effects on corresponding traits in the LJ population. The peak markers for Qtss.saw-4A.1, Qtss.saw-5A.1, and Qgns.saw-4D were not polymorphic between the LJ parents, and thus could not be evaluated. The remaining 10 QTL were evaluated. The effect of Qtss.saw-5D, Qgns.saw-5B.2, Qgns.saw-7A.1, and Qbsss.saw-2B.2 did not differ significantly between the two groups in the LJ population (Figure 1).  Table 3). The additive effects of these QTL on corresponding traits were analyzed based on linked markers. The average corresponding trait values increased as the number of positive alleles increased (Figures 2A-C). Lines with favorable alleles at all the six QTL regions had an average TSS increase of 2.25 vs. those possessing contrasting alleles (Supplementary Table S4, Figure 2A). Lines with both the positive alleles had significantly increased values for BSSS ( Figure 2B). The combination of positive alleles from Qgns.saw-5B.2, Qgns.saw-7A.1, Qgns.saw-4D, and Qgns.saw-1A had the largest effect on GNS (Supplementary Table S4, Figure 2C). Qtss.saw-2B.1, Qtss.saw-2B.2, and Qtss.saw-3B were validated in the LJ population, and the positive alleles of three QTL were derived from Linfen 5064, the additive effects on each corresponding trait were analyzed based on linked markers (Supplementary Table S5, Figure 2D). The combination of positive alleles from Qtss.saw-2B.1, Qtss.saw-2B.2, and Qtss.saw-3B (7.33%, 1.38) had the largest effect on TSS. Compared with lines lacking positive alleles for increased TSS, the positive allele from Qtss.saw-2B.2 significantly increased TSS by 3.30%, which was higher than that for the other single positive alleles of Qtss.saw-2B.1 (1.92%, 0.36) and Qtss.saw-3B (2.45%, 0.46). DH lines with both Qtss.saw-2B.1 and Qtss.saw-3B positive alleles significantly increased TSS (2.56%, 0.48) less than that of DH lines with single positive alleles of Qtss.saw-2B.2 (3.30%, 0.62). These results indicated that the positive allele of Qtss.saw-2B.2 from Linfen 5064 has a larger effect on TSS.

Distribution of Linfen 5064 Favorable Alleles Across Cultivars
The three stable QTL Qtss.saw-2B.1, Qtss.saw-2B.2, and Qtss.saw-3B were detected in more than three environments and were validated in the LJ population. The additive effects of these QTL were from Linfen 5064. Based on the resequencing of 145 wheat cultivars, linkage disequilibrium analysis was performed to assess variation sites within three target QTL regions (Figure 3). Qtss.saw-2B.1, Qtss.saw-2B.2, and Qtss.saw-3B had high recombination rates corresponding to recombination hotspot areas. Therefore, for three QTL the distribution of favorable alleles from Linfen 5064 was analyzed in 145 landmark cultivars ( Table 4). The favorable alleles of Linfen 5064 for Qtss.saw-2B.2 had a lower proportion in the Chinese landraces (CL) (44%) and introduced modern cultivars (IMC) (45%), but a higher proportion in the modern Chinese cultivars (MCC) (77%). Therefore, the favorable alleles of Linfen 5064 at the Qtss.saw-2B.2 locus were selected because of their value in breeding new Chinese cultivars. Qtss.saw-2B.1 and Qtss.saw-3B with the positive Linfen 5064 alleles were less frequent in Chinese landmark cultivars (29.66 and 15.86%, respectively), indicating that Qtss.saw-3B landmark alleles tended to be replaced during breeding by the Linfen 5064 alleles.

Genes Identified in the Major QTL
A series of orthologous GNS-related genes have been cloned in rice (Huang et al., 2009;Kyoko et al., 2009;Qiao et al., 2011;Gao et al., 2016) and wheat (Jiang et al., 2015;Zhang et al., 2015;Shao et al., 2017;Muqaddasi et al., 2019;Rehman et al., 2019), these genes always showed conserved functions across grass species (Valluru et al., 2014). Based on the result of local-blast browse through the IWGSC reference sequence, no homologs of the above genes were found in the physical regions of 690.  Mb on 3BL in wheat. It indicated that there might be novel genes related to GNS among the two QTL, thus, these QTL were chosen for further analysis. Qtss.saw-2B.1 was in the interval 690.21-712.76 Mb on 2BL and where 260 genes have been found in the variety Chinese Spring (CS) (Supplementary Table S6). Gene annotation, expression pattern, and orthologous gene analysis indicate that three genes are likely involved in spike development (Supplementary Table S6, Supplementary Figure S1). The function of TraesCS2B02G500100, TraesCS2B02G500200, and TraesCS2B02G500300 are annotated as a series of molecular signals generated by the binding of the plant hormone abscisic acid to a receptor and ending with modulation of a cellular process. Qtss.saw-3B has 20 genes in CS and 13 common predicated genes between CS and rice (Supplementary Table S7). The genes were not preferentially expressed in spike and grain (Supplementary Figure S2).

New Genes Were Identified in the Interval of the Stable QTL to Control Spike-Related Traits
Genes related to spike traits can be divided into two categories. The first category is flowering time (FT) genes which have significant effects on grain yield, namely, Vrn1, Vrn2/ZCCT1, Vrn3, and Ppd-D1 (Cuthbert et al., 2008;Zhou et al., 2017;Guan et al., 2018). Other genes were mainly involved in spike differentiation which influenced the number of grains per spike by regulating the rate and direction of differentiation. For example, aberrant panicle organization 1 (APO1) controls cell proliferation of the rice meristem, leading to the reduction of the primary and secondary branches of the panicle, thereby affecting panicle development (Kyoko et al., 2009). In addition, some genes can control panicle morphogenesis by regulating hormone and protein expression during rice growth (Huang et al., 2009;Qiao et al., 2011;Gao et al., 2016). BG1 regulates auxin transport and increases biomass, grain number per spike, and grain size to increase yield . In this study, we find three new genes for controlling spike-related traits. TraesCS2B02G500100, TraesCS2B02G500200, and TraesCS2B02G500300 and involved the phytohormone regulatory and ubiquitin proteasoma.
In the next step, we will fine-mapping these QTL which will help explain the formation and development of GNS in wheat and develop linked molecular markers for use by breeders.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
JZhe, WD, JuW, and LQ designed the experiment and developed the original manuscript. LQ, HL, JZha, XZ, JiW, and BW performed the field experiments. LQ, HL, XZ, WD, and JZhe performed the phenotypic data analysis and the QTL detection. WD, JuW, and JZhe revised the manuscript. All authors approved the submitted version of the manuscript.