Epistatic interaction effect between chromosome 1BL (Yr29) and a novel locus on 2AL facilitating resistance to stripe rust in Chinese wheat Changwu 357-9

Four stable QTL for adult plant resistance were identified in wheat line Changwu 357-9, including a new QTL on 2AL showing significant interaction with Yr29 to reduce stripe rust severity. Stripe rust (yellow rust) is a serious disease of bread wheat (Triticum aestivum L.) worldwide. Genetic resistance is considered the most economical, effective and environmentally friendly method to control the disease and to minimize the use of fungicides. The current study focused on characterizing the components of stripe rust resistance and understanding the interactions in Changwu 357-9 (CW357-9)/Avocet S RIL population. A genetic linkage map constructed using a new GenoBaits Wheat 16K Panel and the 660K SNP array had 5104 polymorphic SNP markers spanning 3533.11 cM. Four stable QTL, consistently identified across five environments, were detected on chromosome arms 1BL, 2AL, 3DS, and 6BS in Changwu357-9. The most effective QTL QYrCW357-1BL was Yr29. The 6BS QTL was identified as Yr78, which has been combined with the 1BL QTL in many wheat cultivars and breeding lines. The novel QTL on 2AL with moderate effect showed a stable and significant epistatic interaction with Yr29. The QTL on 3DL should be same as QYrsn.nwafu-3DL and enriches the overall stripe rust resistance gene pool for breeding. Polymorphisms of flanking AQP markers AX-110020417 (for QYrCW357-1BL), AX-110974948 (for QYrCW357-2AL), AX-109466386 (for QYrCW357-3DL), and AX-109995005 (for QYrCW357-6BS) were evaluated in a diversity panel including 225 wheat cultivars and breeding lines. These results suggested that these high-throughput markers could be used to introduce QYrCW357-1BL, QYrCW357-2AL, QYrCW357-3DL, and QYrCW357-6BS into commercial wheat cultivars. Combinations of these genes with other APR QTL should lead to higher levels of stripe rust resistance along with the beneficial effects of multi-disease resistance gene Yr29 on improving resistance to other diseases.


Introduction
Wheat (Triticum aestivum L.) is a major cereal crop consumed widely throughout the world, and production is often constrained by diseases and pests causing substantial yield losses. Stripe rust (yellow rust) caused by Puccinia striiformis Westend. f. sp. tritici Erikss. occurs in almost all wheat-producing regions. This disease causes significant economic losses in terms of reduced grain production and additional costs associated with disease management (Hovmøller et al. 2010;McIntosh et al. 1995). The most profitable and environmentally friendly strategy for farmers to control wheat rusts in both developing and developed countries is to grow genetically resistant wheat varieties (Krattinger et al. 2009).
The application of genomic tools and development of genotyping platforms for wheat improvement have lagged behind other cereals such as rice and maize for a long period largely due to its allohexaploid nature (AABBDD genome), huge genome size (~ 17 Gb) and highly repetitive elements (> 80%). Recent advances in sequencing technology, however, radically changed the landscape and provided opportunities to overcome these difficulties. Over the past few years, great progress was made in developing the reference genome assembly of polyploid wheat and its progenitors, including T. urartu (AA genome) (Ling et al. 2018), Aegilops tauschii (DD genome) (Jia et al. 2013;Luo et al. 2013Luo et al. , 2017Zhao et al. 2017), and wild emmer wheat (T. turgidum ssp. dicoccoides) (AABB genome) (Avni et al. 2017). Subsequent exon capture sequencing and resequencing technologies now accelerate marker development and establish haplotypes associated with resistant and susceptible lines (Cobo et al. 2018;Krasileva et al. 2017;Hao et al. 2020). Single nucleotide polymorphisms (SNPs) as the most abundant and important type of DNA variation were used to develop several high-throughput SNP genotyping platforms such as the 9K, 16K, 35K, 55K, 90K, and 660K high-density SNP chips (Cavanagh et al. 2013;Wang et al. 2014;Jia and Zhao 2016;Qiao et al. 2022). A target sequencing (GBS) system with capture-in-solution (liquid chip) technology known as the wheat 16K SNP array with the advantage of greater power for detection of genetic diversity by linkage disequilibrium decay analysis and genome-wide association studies than the one-amplicon-one-SNP system was developed by a multiple single nucleotide polymorphism (mSNP) approach (Guo et al. 2021).
More than 80 permanently named stripe rust resistance (R) genes (Yr1-Yr83) and many QTLs have been mapped on all 21 wheat chromosomes McIntosh et al. 2017). These genes/QTL can be categorized as all-stage resistance (ASR) and adult plant resistance (APR) or hightemperature adult plant resistance (HTAPR) based on the growth stage at which they can be detected. ASR is often race-specific, qualitatively inherited and controlled by a single gene, whereas APR and HTAPR are more quantitative with individual genes having minor effectiveness, but when combined there are additive effects such that agronomically acceptable levels of resistance are achieved. The added advantage of this type of resistance is durability that is based on the thesis that the genes conferring this type of resistance are non-specific or that any erosion of effectiveness will be a gradual process rather than a 'boom and bust' characteristic of widespread use of single ASR genes (Chen 2005(Chen , 2013Lagudah 2011;Chen and Line 1995), most recently evidenced in China with the emergence of the now-prevalent Yr26-virulent race group, including CYR34. This race group not only overcame Yr26 but also possessed a wide array of virulence for other well-known ASR genes (Wu et al. 2020Huang et al. 2021). Therefore, identification and characterization of APR or HTAPR genes will enrich the overall stripe rust resistance gene pool and thereby accelerate development of wheat cultivars with durable, high-level resistance that can also be combined with effective all-stage resistance (Chen 2013;Liu et al. 2018).
Changwu 131 and Changwu 134 developed by Dr. Zengji Liang (Agricultural and rural Bureau of Changwu County, Xianyang, China) have been commercial wheat cultivars in China for many years. Changwu 357-9 (CW357-9), one of those prefixed as "Changwu" derivatives, has shown a high level of resistance to stripe rust since its release in 1989. However, little was known about the genetic basis of the resistance to stripe rust in this line. The objectives of this study were to: (1) investigate the genetic basis of stripe rust resistance in Changwu 357-9 using a recombinant inbred line (RIL) population tested in multiple environments, (2) identify and map QTL in CW357-9 with significant additive and epistatic effects on resistance to stripe rust using the wheat 16K SNP array, and (3) develop and validate AQP markers closely linked to three identified QTL.

Plant materials
The 167 F 5 -derived F 6 recombinant inbred line (RIL) population was derived from a cross of susceptible Avocet S (AvS) and resistant Changwu 357-9 (CW357-9). A panel of 225 Chinese wheat cultivars/breeding lines and Yr gene carriers were evaluated for response to stripe rust across multiple field environments, and the data were used to determine the prevalence of resistance genes/QTL identified in CW357-9 based on flanking SNP markers . The wheat cultivars Avocet S (AvS), Mingxian 169 (MX169), and Xiaoyan 22 (XY22) were used as susceptible controls.

Greenhouse evaluation
In previous studies, CW357-9 was tested in seedling with Pst races CYR23, CYR29, CYR31, CYR32, CYR33, V26/ CH4 2, V26/Gui22, Su11-4, Su11-5, and Su11-7, and it was susceptible to the currently predominant races CYR32, CYR33, and V26/Gui22 (Wu et al. 2016). In the present study, we used additional three potentially predominant races PST-Lab.1, PST-Lab.2, and PST-V26 collected from field and separated in our laboratory to identify the type of inheritance in CW357-9. The testing regime for seedlings and determination of virulence/avirulence characteristics of PST-Lab.1, PST-Lab.2, and PST-V26 were previously reported in Huang et al. (2021). Infection types (ITs) of all plants were recorded 18 to 21 days after inoculation when the symptoms were fully developed on the susceptible control (AvS and MX169), and based on a 0-9 scale as previously described (Line and Qayoum 1992). The records of IT data were repeated three times to ensure reliability.

Field experiments
The 167 F 6 RILs and parents for disease assessment were grown in five different environments including Jiangyou (JY) in Sichuan province and Yangling (YL) in Shaanxi province during  in Gansu province in 2018-2019, designated as 2018JY, 2019JY, 2018YL, 2019YL, and 2019TS, respectively. Lines carrying Yr29 (Pavon 76, Attila, and Avocet-Yr29) were included as checks. The locations in Sichuan and southern Gansu experience cool, wet weather that is ideal for natural stripe rust survival and spread. At each location, 30 seeds of each line were planted as 1-m single rows and a 30-cm row spacing with a mixture of MX169 and XY22 as susceptible spreaders sown after every 20 rows. Trials at Yangling were inoculated with a mixture of Pst races PST-Lab.1, PST-Lab.2 and PST-V26 suspended in a light oil (1:300) sprayed onto MX169 and XY22 at flag leaf emergence. Two replicates of the RILs were planted in each environment. Stripe rust assessments on adult plants were made 5-25 April at Jiangyou (JY), 3-17 May at Yangling (YL), and 10-15 June at Tianshui (TS), when AvS and XY22 displayed 80% severity or more. Infection types (IT) using a 0 (resistant) to 9 (susceptible) scale (Line and Qayoum 1992) and disease severities (DS) based on the modified Cobb Scale (Peterson et al. 1948) were used to evaluate the adult plant responses to stripe rust. IT and DS of homozygous (not segregated) lines were recorded as single values, and for heterozygous (segregated) lines IT and DS were recorded as two or more values, but later not used in QTL detection. Disease assessment was made at least twice, and the highest IT and DS for each line were used for phenotypic and QTL analyses.

Phenotypic analysis
ANOVA of the mean IT and DS for the RILs in each environment was undertaken to determine the effects of genotype (G), environment (E), and G × E interaction. Pearson's correlation coefficient (r) analysis and ANOVA were conducted using the "AOV" function in QTL IciMapping software 4.1 with the default parameters (Meng et al. 2015). Broad-sense heritabilities (h2 b) of resistance were based on the equation h2 b = σ2 g/(σ2 g + σ2 ge/e + σ2 ε/re), where σ2 g, σ2 ge, and σ2 r represent the genotypic (RILs), G × E and error variances, respectively, and e and r were the numbers of environments and replicates. In addition, the mean phenotypic values for all five environments were used to evaluate the genetic effects and find the best confidence region for each QTL (Mu et al. 2019).

SNP calling and clustering
Genomic DNA were extracted from pools of 10-15 plants from each parent and RIL at the jointing stage using the CTAB protocol (Clarke 2009), and DNA quality was assessed using a NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE, USA). The RILs and parents were genotyped by a new wheat 16K SNP array from Mol Breeding (Shijiazhuang in Hebei province; http:// www. molbr eeding. com). The wheat 660K SNP array from CapitalBio Corporation (Beijing; http:// www. capit albio. com) was used to genotype the two parents. The distribution of SNPs from the 16K array is shown in Table S1. The procedure for marker clustering was described in Huang et al. (2021).

Linkage map construction and QTL analysis
A Chi-squared (χ2) test for goodness of fit to a 1:1 segregation ratio was performed for each SNP before processing by including those < 10% missing values and major allele frequencies (MAF) ≤ 95%. One marker was selected from each co-segregating marker group using the "BIN" function in IciMapping V4.2 software. The selected markers were used to generate the genetic map using the "MAP" function in IciMapping V4.2 software and drawn in Mapchart V2.3 (Meng et al. 2015;Voorrips 2002). Combining the calculated value by 1000 permutations at a probability of 0.01, the logarithm of odds (LOD) score to determine significant QTL was 6.8 in all five environments. Recombination fractions were converted to cM using the Kosambi function (Kosambi 1943). The phenotypic data including IT, DS, and mean values from all environments were used to identify the QTL. Inclusive composite interval mapping with the additive tool (ICIM-ADD) in IciMapping V4.2 was performed to detect QTL. The phenotypic variances explained (PVE) by individual QTL and additive effects at the LOD peaks were also obtained. Due to low marker density, some QTL mapped in potentially large regions. To further narrow down the flanking intervals of target loci, significant SNPs from 660K SNP array were converted into allele-specific quantitative PCR (AQP) markers by JasonGen Biological Technology Co., Ltd (Beijing; http:// www. jason gen. com) to genotype the RIL population.

Epistasis
Genotyped SNP markers associated with stripe rust resistance across five environments, were used for pairwise interaction analysis in Network version 2.1 (Yang et al. 2008). QTL effects were evaluated by the mixed linear model (MLM) approach. A ''2D genome scan'' option was used to map epistatic QTL with or without single-locus effects. additive × additive (A*A) epistatic effects of mapped using the ''map epistasis'' function. F values were used to control the error rate by permutation tests.

Genetic linkage map
Of the 20,995 SNPs, 5104 (24.4%) showed polymorphism between the parents. By using the "BIN" function in QTL IciMapping 4.2, redundant polymorphic SNPs were removed showing > 10% missing data and distorted segregation. Finally, 841 SNPs were chosen to construct the genetic linkage map; they were distributed in 22 linkage groups spanning 3533.11 cM. The A, B, and D genomes included 290 (34.48%), 374 (44.47%), and 177 (21.05%) markers covering lengths of 1268.96, 1356.53, and 1012.27 cM with average marker intervals of 4.38, 3.63, and 5.72 cM, respectively. Only chromosome 2D had two linkage groups; the other chromosomes were each represented by a single linkage group (Table S1).

Phenotypic evaluation
The CW357-9 seedlings were resistant (IT 3-4) to PST-Lab.1 and PST-Lab.2, but susceptible (IT 8-9) to PST-V26. CW357-9 was highly resistant (IT 1-2, DS ≤ 5%) at the adult plant stage in the field, whereas AvS was highly susceptible (IT 8-9) in all experiments. Based on these results, CW357-9 possessed both seedling resistance to two potentially predominant races PST-Lab.1 and PST-Lab.2 and APR in the field. In the seedling test, all the RILs and two parents were phenotyped by PST-Lab.1 to identified ASR. In the field experiments, both IT and DS data for RILs showed continuous distributions (Fig. 1), indicating that resistance in CW357-9 was quantitatively inherited. Pearson's correlation coefficients of pairwise comparisons of IT and DS ranged from 0.60-0.85 and 0.58-0.88 (P < 0.001) (Table 1), respectively. Broad-sense heritabilities for both IT and DS were 0.92 (Table 2). P values in the ANOVA for IT and DS were highly significant (P < 0.0001) for RILs, environments, and line × environment interactions. Lack of significant variation between the replicates suggested that genetically controlled resistance was the main source of phenotypic variation in the RIL population (Table 2). These results indicated that the QTL conferring resistance was effective in the five environments.

Epistatic interaction detected by QTL Network version 2.1
Significant epistatic interactions were detected across all field traits using QTL Network version 2.1. Two different intervals on 1BL and 2AL corresponding to the markers

QTL combinations and interaction
In order to investigate the effects of QTL combinations, RILs were classified into five genotypic groups based on all field tests (Table S2). RILs with all four QTL QYrCW357-1BL, QYrCW357-2AL, QYrCW357-3DL, and QYrCW357-6BS were more resistant (lower IT and DS) than all others, displaying resistance similar to CW357-9 (Fig. 3a, b; Table S2). Among these genes. the combination of QYrCW357-1BL and QYrCW357-2AL showed the most significant effect in reducing stripe rust severity (Fig. 3c, d). RILs with none of the four stable QTLs had mean IT and DS values of 8.2 and 88.6%, respectively; RILs with only one QTL (1BL or 2AL) had mean values of 6.9 and 71.0% for the 1BL locus (similar to Avocet-Yr29 in Table S3) and 7.2 and 76.1% for 2AL, respectively (Fig. 3c, d). The group combining QYrCW357-1BL and QYrCW357-2AL, with mean IT and DS of 4.6 and 36.0%, respectively, showed significant effect in reducing IT and DS (Fig. 3c-f).

Polymorphisms of AQP markers for stripe rust resistance in wheat genotypes
To determine the robustness of identified markers for stripe rust resistance in CW357-9, genotyping of the 225-accession panel for polymorphic AQP markers AX-110020417, AX-110974948, AX-109466386 and AX-109995005 represented for QYrCW357-1BL, QYrCW357-2AL, QYrCW357-3DL, and QYrCW357-6BS, respectively, suggested these markers were significantly associated with the DS scores of the wheat panel (Table S4). The genotyping assays generated three groups for different combination, enabled by testing the user-friendly markers for validation of both epistatic and additive effects (Fig. 4). Wheat lines with both QYrCW357-1BL (or Yr29) and QYrCW357-2AL were on the average more resistant than lines without them, but some accessions containing the QTL were highly susceptible, indicating that the effects of the two QTL alone could be influenced by recombination between markers, genetic background, and environment. However, wheat lines combining all four loci had the lowest average DS in Yangling and Tianshui (Fig. 4). Sequences for the AQP markers AX-110020417,  Table S4.

Discussion
There is now strong evidence that pyramiding multiple partially effective resistance genes with additive or positive interaction in a single wheat cultivar can lead to more durable resistance than a single highly effective all-stage resistance gene . The data also suggest that the level of resistance required to protect yield potential and to prevent significant disease spread will require about four genes Zeng et al. 2019). In addition, the numbers of epistatic interactions are frequently larger than the number of additive QTL, and the importance and number of epistatic interactions in terms of both the number of loci involved and effects may be greater than the additive QTL (Malmberg et al. 2005;Liu et al. 2022). CW357-9 is such a wheat genotype combining four partial APR QTL with both epistatic and additive effects and has maintained highly resistance for more than ten years in China.

Yr29 and Yr78 are frequently combined in Chinese wheat germplasm
Combinations of QYrCW357-1BL (Yr29) on chromosome arm 1BL and QYrCW357-6BS (Yr78) on 6BS were detected in several Shaanxi wheat cultivars, including Qinnong 142 , Shaanong 33 , and Xinong3517 (Huang unpublished data). These cultivars were highly resistant in the field at Shaanxi, Gansu, and Sichuan provinces, which are hotspot regions for over-season survival of Pst and have frequent occurrence of stripe rust. In this study, most carrier varieties with Yr29 and Yr78 were from Sichuan (12, 21.4%), Henan (9, 16.1%), and Shandong (8, 14.3%). Similar results reported in Huang et al. (2021) indicate varieties carrying Yr29 and Yr78 are common in these provinces.

Gene-gene interaction contributing to stripe rust resistance
A method of MAS based on QTLs with epistatic effects was proposed (Liu et al. 2003). Changwu 357-9 with desirable agronomic traits and a high level of durable resistance to stripe rust can be used as a parent for marker-assisted breeding for favorable epistatic interactions. Based on the epistatic analysis of field IT and DS, QYrCW357-1BL (or Yr29) and QYrCW357-2AL showed a significant interaction (Table 4, Fig. 2). Yr29 is present in many wheat cultivars around the world and has remained effective for more than 60 years (Cobo et al. 2017). The novel locus on chromosome arm 2AL interacted with Yr29 and other genes to confer an acceptable level of resistance to stripe rust in Chinese wheat Changwu 357-9 ( Fig. 2c-f; Table S2). Based on genotyping of the flanking AQP markers, these results suggested that these markers can be used for developing new cultivars with high-level of durable resistance to stripe rust (Table S4). In addition, further exploration may provide insight for understanding the interactions observed between QYrCW357-1BL or Yr29 and QYrCW357-2AL in this study as well as functional mechanisms that contribute to this resistance gene network.

Conclusion
CW357-9 with durable resistance to stripe rust for more than a decade carries a 4-gene combination of APR genes, including Yr29, Yr78, QYrCW357-2AL, and QYrCW357-3DL with additive and epistatic effects. The QTL on chromosome arms 2AL and 3DL were novel. The key points from 1 3 this work were: (1) QYrCW357-2AL and QYrCW357-3DL can be selected to enrich the overall stripe rust resistance gene pool for breeding; (2) the combination of Yr78 and Yr29 is frequent among wheat cultivars and breeding lines in China; and (3) the discovery of favorable epistatic interaction between Yr29 and QYrCW357-2AL. Finally, CW357-9 not only represents a useful breeding parent but the markers developed here can be potentially used in MAS to develop new cultivars with potentially durable resistance. Field trials in disease nurseries will still be required to determine that lines with the selected resistance gene combination confer an acceptable level of protection from stripe rust.