Two QTL for kernel number per spike identified from durum wheat

Abstract As one of the three yield components, kernel number per spike (KNPS) plays a key role in determining wheat yield. In this study, in combination with the previously constructed genetic map for a tetraploid wheat recombinant inbred line (RIL) population developed from the cross between a durum wheat and a wild emmer wheat and phenotype data from four environments, two major and stably expressed quantitative trait loci (QTL) for KNPS were identified. The phenotypic variation explained by these two QTL, Qknps-DW-3A.1 and Qknps-DW-3A.2, was up to 12.52% and 29.52%, respectively. The positive alleles were both from the durum wheat. Pyramiding analysis suggested that the combination of these two positive alleles has the largest effect on increasing KNPS. Comparison of physical interval for these two QTL with those reported previously showed that they may be novel loci. Genetic correlations between KNPS and other agronomic traits were also evaluated. Taken together, the two major and stably expressed QTL for KNPS from tetraploid wheat reported here they should be useful in wheat breeding.


Introduction
Bread wheat (Triticum aestivum L.) provides about 20% of food energy to humans (http://www.fao.org/faostat), and thus, it is important for global food security. Wheat yield is determined by the spike number per square metre, kernel number per spike (KNPS) and 1000-grain weight. Of these, KNPS was regarded as a significant factor increasing wheat yield, as it has been steadily improved [1,2]. It is thus an urgent work to discover crucial QTL for KNPS for wheat breeding programs.
To date, genes or QTL associated with KNPS have been identified on all the 21 chromosomes of bread wheat [3][4][5][6][7][8][9][10][11][12][13][14][15][16]. Researchers have suggested that improvement of wheat yield and quality was impeded due to its narrow genetic basis [17]. It is thus essential to extend the wheat gene pool and identify novel potential genes or QTL to improve wheat breeding. Fortunately, wheat related species carry desirable genes that can be utilised for wheat improvement. For example [18], detected a locus for grain number per spike from the wheat-Agropyron cristatum 6 P disomic addition line. Recently [19], cloned Pm41, a powdery mildew resistance gene derived from wild emmer wheat.
Recently, QTL for effective tiller number (ET) from a tetraploid wheat recombinant inbred line (RIL) population developed from the cross between a durum and a wild emmer accession were identified [29]. Here, we attempt to excavate QTL for KNPS from this reported RIL population aiming at providing gene sources for wheat improvement.

Population and field experiment
The RIL population used in this study developed by single seed descent from an initial F 2 population was derived from a cross between a durum wheat (DU09123, Triticum turgidum L. var durum) and a wild emmer genotype (WE34021, T. turgidum subsp. dicoccoides) and it included 116 F 6 -F 9 lines [29]. The agronomic traits of the RILs and their parents were measured in four consecutive years, 2017-18 (E1) and 2018-19 (E2), 2019-20 (E3), 2020-21 (E4) at the experimental field of Southwest University of Science and Technology (MianYang, Sichuan, China). Field management followed local practices. Lines were planted in a single row at a density of 15 kernels/150 cm row, and a row spacing of 30 cm. Agronomic traits from five plants for a given line including plant height (PH), spikelet number per spike (SNS), spike length (SL) and ET, flowering date (FD), KNPS, kernel weight per spike (KWS), kernel length (KL) and width (KW) and thousand kernel weight (TKW) were evaluated according to the previous study [29]. Specifically, PH was measured from the plant base to the tip of the spike (excluding awns). SNS and SL was determined from the main spike of a plant. ET indicated the effective spikes of a single plant. FD was calculated from the sowing date to that when half of the plants for a given line flowered. KNPS and KWS from five representative and well-pollinated spikes per line were counted and measured. After threshing and drying, the kernels were used for measuring kernel length (KL), width (KW) and thousand kernel weight (TKW).

Data analysis
Pearson's correlation, frequency distribution, standard error and the Student's t-test (p < 0.05) were used in the IBM SPSS Statistics 25 (Armonk, NY, USA). The broad-sense heritability (H 2 ) and best linear unbiased prediction (BLUP) of agronomic traits from multiple environments were calculated using SAS version 8.0 (Cary, North Carolina).

QTL detection
The previously constructed genetic linkage map using the wheat 55 K single nucleotide polymorphism (SNP) array [29] was adopted for QTL analysis. The map included 1154 bin markers mapped on 14 linkage groups with an average density of 2.13 cM/marker. QTL analysis was carried out as described previously [29]. Individual environment QTL identification was carried out using the biparental populations (BIP) module in IciMaing4.1. The QTL-by-environment interactions were analysed using the function of multi-environment trial module in IciMaing4.1. QTL that were identified in at least two environments and explained more than 10% of the phenotypic variation were regarded as major ones, while those having common flanking markers were treated as a single one.

Predicted genes in the interval of QTL
We used the sequences of the SNP markers to blast against the wild emmer (accession Zavitan) [30], durum wheat (accession Svevo) [31] and bread wheat genotype (accession 'Chinese Spring' , CS) (IWGSC RefSeq v2.1) [32] genome sequences to identify their physical locations to determine the physical interval of the QTL. Predicted genes located between the flanking markers and their function annotations were retrieved from the Triticeae Multi-omics Centre [33].  (Figure 1). There were significant correlations for KNPS among the four environments with coefficients ranging from 0.45 to 0.98 (p < 0.01, Table 2).

QTL detection
Using QTL mapping analysis based on the linkage map and phenotype data across four environments and the BLUP dataset, 13 QTL for KNPS were identified in the single-environment analysis. They were mapped on chromosomes 1 A, 3 A (2 loci), 6 A (4), 7 A, 1B, 5B (4) and 7B. The LOD value for these QTL ranged from 2.94 to 39.83. The phenotype variance explained (PVE) ranged from 1.51 to 29.52%. The positive alleles of these QTL were all from DU (Table 3).
Qknps-DW-3A.1 was stably expressed in two of the four environments and the BLUP dataset, explaining 10.05 to 12.52% of the phenotypic variation. The LOD score ranged from 4.45 to 6.76. It was thus regarded as a major QTL (Figure 2 and Table 4 Table S1).
Qknps-DW-3A.2 was stably detected in three environments and the BLUP dataset, explaining 13.54 to 29.52% of the phenotypic variation. The LOD score ranged from 3.64 to 39.83. It was also regarded as a major QTL ( Figure. 2 and Table 3    3AL of CS genome. There were 1 and 6 predicted genes in the respective intervals with 1 ortholog in the corresponding intervals (Supplemental Table S2).
In addition, Qknps-DW-6A.1 and Qknps-DW-5B.1 can be detected in two environments, with phenotypic variation explained up to 9.16 to 12.49%, respectively. We also detected Qknps-DW-6A.2, Qknps-DW-6A.4, Qknps-DW-7A and Qknps-DW-5B.3, the PVE of which was larger than 10%. However, as they were all identified in only a single environment, we think they were not stably expressed. We thus excluded them from further analysis.
A total of 21 QTL were detected using the QTL-byenvironment interaction analysis (Supplemental Table  S3). Of them, Qknps-DW-3A.1 and Qknps-DW-3A.2 were also detected in individual environment analysis, indicating that they were major QTL for KNPS.

Effects of the two major QTL on KNPS
The effects of the two major QTL Qknps-DW-3A.

Correlations between SNS and the two major QTL with other agronomic traits
The phenotypic correlations between SNS and other agronomic traits were analysed and listed in Table 4. KNPS was positively correlated to FD, SNS and KWP, and negatively correlated to KL (p < 0.01). Significant correlation was not detected between KNPS and PH, ET, SL, KW and TKW (Table 4).
To further analyse the effects of the two major QTL on other agronomic traits, the 114 lines were grouped into two classes (excluding heterozygous lines) based on the genotypes of the flanking markers. For Qknps-DW-3A.1, 53 lines carrying homozygous alleles from WE showed no significant difference with 51 ones carrying those from DU in PH, FD, ET, SL, SNS, KL, KW, TKW and KWS ( Figure S1). For Qknps-DW-3A.2, 55 lines with homozygous alleles from WE also had no significant difference with 54 ones with those from DU for the measured traits (Supplemental Figure S2).

Discussion
Excavating, characterisation and mapping loci for yield-related traits from wheat related species can provide important resources for bread wheat improvement. In this study, the wheat 55 K SNP array based constructed linkage map in combination with four environments were used to map two major QTL for KNPS, the positive alleles of which were both from the durum wheat DU accession. These two QTL were not likely interacted with FD and PH. Pyramiding analysis further suggested their great potential in wheat breeding. Although numerous QTL for KNPS have been reported on nearly 21 chromosomes of bread wheat, only several QTL were identified on chromosome 3 A and further very few stable and major QTL for KNPS were identified. For example [2], detected three KNPS-associated loci on bread wheat 3 A chromosome through a genome-wide association analysis of grain number using 531 SSR markers. Of them, Xgwm2-3A (~60.58 Mb on 3 A) was detected in two of the five environments, and Xcfa2234 (~270.76 Mb on 3 A) and Xgwm162 (~751.92 Mb on 3 A) were detected in only a single environment [16] found that QKNPS-DH-3A-1.1 (~488.87 Mb) and QKNPS-DH-3A-2.2 (~184.50 Mb) were detected on bread wheat 3 A chromosome in only one of the five environments. The previously reported QTL were mainly identified in bread wheat and they were physically far away from Qknps-DW-3A.1 and Qknps-DW-3A.2 (~ 650-740 Mbp) identified in this study. Additionally, the favourable alleles of the presently identified QTL were from durum wheat. These results suggested that Qknps-DW-3A.1 and Qknps-DW-3A.2 might be novel loci and should be focussed on in wheat breeding. Here, no significant effects of the two QTL for KNPS on PH, FD and other agronomic traits were detected, suggesting that these two QTL may have functional roles independent of PH and FD. Previous reports have shown that pyramiding of favourable genes is a good way to improve yield [34][35][36]. Our results showed that the combination of positive alleles of both Qknps-DW-3A.1 and Qknps-DW-3A.2 had the largest effect on KNPS. These results indicate the great potential of pyramiding both QTL to regulate KNPS and thus improve wheat yield. Further, the predicted candidate genes in the intervals where major QTL were located will be useful for future fine mapping.

Conclusions
The KNPS for the mapping population was measured in four environments. Correlations between KNPS and other agronomic traits were evaluated. Using the constructed linkage map based on the wheat 55 K SNP, two major and stably expressed QTL for KNPS were detected on chromosome arm 3AL. Their positive alleles were both from durum wheat. Physical maps analysis indicated that these two QTL may be novel loci for KNPS. Genetic analysis suggested that they were not correlated to PH and FD. Pyramiding analysis further showed the great breeding potential of these two QTL given their largest additive effect.

Disclosure statement
No potential conflict of interest was reported by the authors.

Data availability statement
All data generated or analysed during this study are included in this article and its supplementary information files; further inquiries can be directed to the corresponding author.

Funding
This study was supported by The Open Project from the Ecological Security and Protection Key Laboratory of Sichuan Province (ESP1608,ESP1801), The Scientific research initiation project of Mian yang normal university (QD2019A13).