Identification of QTLs for Stripe Rust Resistance in a Recombinant Inbred Line Population

Stripe rust, caused by Puccinia striiformis f. sp. tritici (Pst), is one of the most devastating fungal diseases of wheat worldwide. It is essential to discover more sources of stripe rust resistance genes for wheat breeding programs. Specific locus amplified fragment sequencing (SLAF-seq) is a powerful tool for the construction of high-density genetic maps. In this study, a set of 200 recombinant inbred lines (RILs) derived from a cross between wheat cultivars Chuanmai 42 (CH42) and Chuanmai 55 (CH55) was used to construct a high-density genetic map and to identify quantitative trait loci (QTLs) for stripe rust resistance using SLAF-seq technology. A genetic map of 2828.51 cM, including 21 linkage groups, contained 6732 single nucleotide polymorphism markers (SNP). Resistance QTLs were identified on chromosomes 1B, 2A, and 7B; Qyr.saas-7B was derived from CH42, whereas Qyr.saas-1B and Qyr.saas-2A were from CH55. The physical location of Qyr.saas-1B, which explained 6.24–34.22% of the phenotypic variation, overlapped with the resistance gene Yr29. Qyr.saas-7B accounted for up to 20.64% of the phenotypic variation. Qyr.saas-2A, a minor QTL, was found to be a likely new stripe rust resistance locus. A significant additive effect was observed when all three QTLs were combined. The combined resistance genes could be of value in breeding wheat for stripe rust resistance.


Introduction
Bread wheat (Triticum aestivum) is one of the most important food crops for mankind and the security of wheat production benefits economic development and social stability. However, wheat production in China is continually challenged by diseases, including rusts, powdery mildew, and Fusarium head blight. Stripe rust, caused by Puccinia striiformis f. sp. tritici (Pst), is one of the most devastating fungal diseases in many areas around the world. Beddow et al. [1] estimated that up to 88% of the world's wheat cultivars had become susceptible since 1960 and that annual losses amounted to 5.47 million tonnes. Resistance is recognized as the most effective, economic, and environmentally safe strategy for control of stripe rust, although fungicides can also effectively control the disease, provided they are used in a timely and safe manner [2,3].
Resistance to stripe rust is generally categorized as seeding (or all-stage) resistance and adult-plant resistance (APR, including high temperature APR) according to the growth stage at which it is expressed [2,4]. Up until now, seventy-nine genes for stripe rust resistance (Yr1 to Yr79) have been permanently named, but dozens of temporarily designated and hundreds of quantitative trait loci (QTL) have been reported and mapped to the wheat genome [5,6]. Among the formally designated

Phenotypic Analysis
The frequency distributions of disease severities for stripe rust reaction at adult plant stages in each environment ranged over 0-100, 0-95, 2.5-100, and 0.5-100 at XD2016, XD2017, JT2017, and XC2017, respectively, showing continuous variation ( Figure 2). The results indicated that stripe rust resistance in CH55 and CH42 was possibly controlled by multiple genes. In all four environments there was transgressive segregation in both directions (Table 1). Broad sense heritability (H 2 ) was 0.88 (Table 1), indicating the data could be used for further QTL mapping.

Phenotypic Analysis
The frequency distributions of disease severities for stripe rust reaction at adult plant stages in each environment ranged over 0-100, 0-95, 2.5-100, and 0.5-100 at XD2016, XD2017, JT2017, and XC2017, respectively, showing continuous variation ( Figure 2). The results indicated that stripe rust resistance in CH55 and CH42 was possibly controlled by multiple genes. In all four environments there was transgressive segregation in both directions (Table 1). Broad sense heritability (H 2 ) was 0.88 (Table 1), indicating the data could be used for further QTL mapping.

Analysis of SLAF-Seq Data and SNP Markers
After SLAF library construction and sequencing, 354.218 Gb of data containing 1771.45 M pairedend reads were obtained; 94.82% of the bases were of high quality with Q30 (a quality score of 30 indicates a 1% chance of error, thus a 99% accuracy) and the guanine-cytosine (GC) content was 44.79% (Table 2). A total 2,825,198 SLAFs were developed. The SLAFs numbers for CH42 and CH55 were 862,053 and 863,835, and their corresponding average sequencing depths were 26.70 and 24.15, respectively. The average number of SLAFs for the RIL population was 493,537 and the average sequencing depth was 11.80 (Table 3). The range of reads in the RILs was 2,422,631 to 14,454,545.

Analysis of SLAF-Seq Data and SNP Markers
After SLAF library construction and sequencing, 354.218 Gb of data containing 1771.45 M paired-end reads were obtained; 94.82% of the bases were of high quality with Q30 (a quality score of 30 indicates a 1% chance of error, thus a 99% accuracy) and the guanine-cytosine (GC) content was 44.79% (Table 2). A total 2,825,198 SLAFs were developed. The SLAFs numbers for CH42 and CH55 were 862,053 and 863,835, and their corresponding average sequencing depths were 26.70 and 24.15, respectively. The average number of SLAFs for the RIL population was 493,537 and the average sequencing depth was 11.80 (Table 3). The range of reads in the RILs was 2,422,631 to 14,454,545.  Among those, the markers in which the bases were absent in the paternal or maternal parent were filtered, leaving 640,734 markers. Then, markers with average sequence depths < 4 were filtered, leaving 446,616 markers. Among the 446,616 SNP markers, 162,394 markers were polymorphic SNPs with a polymorphic rate of 36.36%. All the polymorphic SNP markers were classified into four genotypes-aa × bb, hk × hk, lm × ll, and nn × np. However, only the genotype aa × bb, consisting of 75,347 SNP markers, which accounted for 3.01% of the total SNP markers, was used for further analysis. Finally, markers with parental sequence depths of less than ten and significant segregation distortions of less than 0.01 (p < 0.01) were filtered and the remaining 6732 markers were used to construct the final genetic map (Table 4).

Genetic Map Construction and Consistency Analysis
The genetic map of 21 linkage groups was 2828.51 cM, with an average marker interval of 0.42 cM. The sub-genome statistics are provided in Table 5. The largest chromosome was chromosome 7B and the shortest was chromosome 6D. The largest gap in the map was 19.46 cM, which was located on chromosome 2B. The largest proportion of gaps less than 5 cM was 99.77% for chromosome 6A, whereas the smallest proportion of gaps, 93.85%, was for chromosome 4D ( Figure 3).    The consistency analysis of the SNP loci between the genetic map and the physical map is shown in Table 6. Among all 21 linkage groups, there were 10 chromosomes with Spearman coefficients between 0.8 and 0.9, including 1A, 2D, 3A, 4B, 4D, 5B, 5D, 6A, 6D, and 7A. The Spearman coefficients for the other 11 chromosomes were between 0.9 and 1.0. The results indicated that the locations of most SNP loci on the genetic map were consistent with their corresponding physical locations in the Chinese Spring genome.
Qyr.saas-1B and Qyr.saas-7B were detected in all four environments, explaining 6.24%-34.22% and 3.27-20.64% of the phenotypic variation, respectively. Qyr.saas-1B was located in a 13.05 cM interval flanked by markers 90327 from XD2017 and 90695 from XD2016 and Qyr.saas-7B was located in a 13.16 cM interval flanked by markers 66151 from XD2017 and 66313 from XD2016. Qyr.saas-2A, with a smaller effect, was detected in only two environments and was located in a 12.12 cM region flanked by markers 71619 from XD2017and 71915 from JT2017, explaining 3.77-5.29% of the phenotypic variation.

Additive Effects of QTLs
The RILs were divided into 8 groups based on the genotypes of the closest markers for each of the three QTLs (Table 8). Significant additive effects were found in RILs with two or more resistance QTLs. Qyr.saas-1B and Qyr.saas-7B significantly reduced disease severity in all four environments when present alone. When present together in XC2017, they acted in an additive fashion and conferred lower severity than either QTL alone. A similar additive effect occurred in JT2017 in combinations of Qyr.saas-1B and Qyr.saas-2A. The lowest severities occurred when all three QTLs were combined in XD2016; the disease severities of many RILs approached immunity. Extremely low disease severity scores also occurred in XD2017 and XC2017, but were not apparent in JT2017.

Additive Effects of QTLs
The RILs were divided into 8 groups based on the genotypes of the closest markers for each of the three QTLs (

Additive Effects of QTLs
The RILs were divided into 8 groups based on the genotypes of the closest markers for each of the three QTLs (

Additive Effects of QTLs
The RILs were divided into 8 groups based on the genotypes of the closest markers for each of the three QTLs (

Additive Effects of QTLs
The RILs were divided into 8 groups based on the genotypes of the closest markers for each of the three QTLs (Table 8). Significant additive effects were found in RILs with two or more resistance QTLs. Qyr.saas-1B and Qyr.saas-7B significantly reduced disease severity in all four environments when present alone. When present together in XC2017, they acted in an additive fashion and conferred lower severity than either QTL alone. A similar additive effect occurred in JT2017 in combinations of Qyr.saas-1B and Qyr.saas-2A. The lowest severities occurred when all three QTLs were combined in XD2016; the disease severities of many RILs approached immunity. Extremely low disease severity scores also occurred in XD2017 and XC2017, but were not apparent in JT2017. The same letter within a column indicates no significant difference at p > 0.05. "+" means containing the corresponding QTL while "-" means no QTL.

Discussion
In order to determine the relationship between the three QTLs identified in the present study and other Yr genes and QTLs reported previously, we compared their physical locations by basic local alignment search tool (BLAST) analysis of the International Wheat Genome Sequencing Consortium (IWGSC) RefSeq v1.0 genome, which was shown in Figure 4 and Table 9.
Yr1 [41], Yr32 [42], and YrJ22 [43] were mapped to 2AL, but these were genes of large effect. Qyr.saas-2A was likely to be a new stripe rust resistance locus based on its different physical location (Table 9, Figure 4b). This QTL was detected only in XD2017 and JT2017, which explained 3.77% and 5.29% of the phenotypic variation, respectively ( Table 7). The effect of Qyr.saas-2A was much smaller than that of Qyr.saas-1B and Qyr.saas-7B. However, a significant additive effect in Qyr.saas-2A was observed. Singh et al. [44] indicated that an adequate level of slow rusting resistance could be achieved by the additive/complementary effects of three to five genes. This has been supported by many reports, including those of Yang et al. [45], Lan et al. [24], and Rosewarne et al. [46]. Similarly, the disease severities of RILs approached immunity when Qyr.saas-2A was combined with two other QTLs, Qyr.saas-1B and Qyr.saas-7B, in XD2016. There is repeated evidence that an effective and stable level of adult plant stripe rust resistance can be achieved by using combinations of genes that individually confer relatively small effects. Therefore, although the effect of Qyr.saas-2A was small, it provided enhancement effects and therefore could be useful in Sichuan wheat breeding for multiple gene resistance to stripe rust.
Based on the studies of chromosome composition of CH55 revealed by FISH analysis (Figure 1), we found that CH55 contained both 1RS·1BL and 5B-7B reciprocal translocation chromosomes. The 1RS·1BL translocation is still widely used in wheat breeding because of the superior genes for grain yield and stress tolerances in 1RS [47]. In the present study, the 1BL arm of 1RS·1BL in CH55 carried the stripe rust resistance QTL Qyr.saas-1B. According to a previous study, the alien chromatin suppresses the recombination between normal and translocated chromosomes [48]. Therefore, the selection of 1RS·1BL accumulates excellent agronomic characteristics and resistance with high frequency in breeding practice. Moreover, the 5B-7B reciprocal translocation is possibly of French origin according to the genealogy of CH55. It was found that the stripe rust resistance QTL QYr.ufs-5B was located on 5BS in the 5B-7B reciprocal translocation [49], which requires further validation by Pst races in different environments for CH55.

Plant Materials and Field Trials
A set of 200 F 6 RILs developed from cross CH42/CH55 and parents were used to evaluate stripe rust responses in multiple environments. CH42 and CH55, developed by Crop Research Institute, Sichuan Academy of Agricultural Sciences, were released in 2004 and 2009, respectively. CH42 is a synthetic wheat derivative produced from cross SynCD768/SW3243//Chuan 6415 and CH55 was selected from the cross SW3243/SW8688. Chuanyu12, developed by Chengdu Institute of Biology of Chinese Academy of Sciences, is highly susceptible to the currently prevalent Pst races in Sichuan province and was used as spreader.
Field trials were conducted at the Xindu (Sichuan Province) research station in the 2015-2016 (XD2016) and 2016-2017 (XD2017) growing seasons and also at Jitian and Xichong (Sichuan Province) in 2016-2017 (JT2017, XC2017). Field trials were conducted in randomized complete blocks with two replications. Plots were sown as single 1 m rows, 25 cm apart, and about 30 seeds were sown in each row. Every 20th row was planted with the susceptible cultivar Chuanyu 12 as a spreader to produce an inoculum and as a control. The surrounding spreaders were inoculated with a mixture of currently prevalent Pst races, including CYR33, CYR34(V26), and G22-14. Adult-plant disease severities were visually recorded as 0-100% according to the modified Cobb Scale of Peterson et al. [50] when severities on Chuanyu 12 reached 90-100%, usually 1-2 weeks post-anthesis.

Broadsense Heritability
Phenotypic variance per plot in multi-trials can be written as σ 2 P = σ 2 G + σ 2 GE + σ 2 ε , where σ 2 G is the genetic variance, σ 2 GE is the variance for genotype-environment interaction, and σ 2 ε is the residual variation. Broad sense heritability in on an individual plot basis was calculated with the formula

DNA Extractions
Young leaf tissues of 1 plant per parent and RILs were sampled in 2016, stored at −80 • C, and used for DNA extraction. The genomic DNA from each genotype was extracted using the cetyltrimethylammonium ammonium bromide (CTAB) method.

SLAF Library Construction and High-Throughput Sequencing
In this project, we used the wheat reference genome version IWGSC1.0 downloaded from ftp://ftp.ensemblgenomes.org/pub/plants/release-30/fasta/triticum_aestivum/dna/. SLAF-seq was used to genotype the 200 RILs and parents using the procedure designed by Sun et al. [14], with minor modifications. DNA was digested into 464-484 bp fragments using restriction enzyme RsaI. The digested fragments were modified by adding nucleotide A and Dual-index sequencing adapters were ligated to the A-tailed DNA, which was amplified and purified to the target fragments. The purified fragments were sequenced on an Illumina HiSeq TM platform.

Analysis of SLAF-Seq Data and Genetic Map Construction
SLAF marker identification and genotyping were performed following Sun et al. [14]. SNP loci in each SLAF locus were detected using the genome analysis toolkit (GATK) software. A genetic map was constructed for filtered markers using HighMap software and referring to the procedure detailed by Zhang et al. [51]. Spearman coefficients were used to analyze the consistency between the genetic and physical maps.

QTL Analysis
QTL IciMapping V4.0 [52,53] was used to identify QTLs by ICIM. A LOD score of 3.0 was used as a threshold for the declaration of linkage and the Kosambi mapping function was used to convert recombination frequencies into map distances. If a QTL was detected in a single environment, it was discarded. A positive additive effect indicated that the favorable allele was from the CH55 parent, whereas a negative additive effect indicated that the favorable allele was from CH42.

Conclusions
This study constructed a high-density genetic map and identified three QTLs for stripe rust resistance using the CH42/CH55 RILs via SLAF-seq technology. The genetic map of 21 linkage groups was 2828.51 cM, with an average marker interval of 0.42 cM. Qyr.saas-7B was derived from CH42, whereas Qyr.saas-1B and Qyr.saas-2A were from CH55. Qyr.saas-2A was likely to be a new stripe rust resistance locus. A significant additive effect was observed when all three QTL were combined. The combined resistance genes could be of value in breeding wheat for stripe rust resistance.