Genetic study of the resistance of faba bean ( Vicia faba ) against the fungus Ascochyta fabae through a genome- wide association analysis

Ascochyta fabae is a fungal pathogen responsible for marked yield losses in spring and winter faba beans worldwide. The aim of this genome- wide association study (GWAS) using 188 diverse winter faba bean inbred lines was to exploit earlier Ascochyta blight resistance studies and to identify new resistance loci. Phenotyping after artificial inoculation under controlled conditions revealed significant variation for all eight scored disease traits. This GWAS was based on 1829 AFLP marker and 229 SNP marker, including 17 so- called ‘guide’ SNP markers. The latter were identi fied by map fragment alignments between the consensus smap of Webb et al., (2016, Plant Biotechnology Journal, 14, 177– 185) and three earlier published Ascochyta blight resistance studies. A total of 12 markers were found significantly associated with six traits, explaining 5.6% to 21.7% of the phenotypic variance. One ‘guide’ SNP

They suggested the presence of six QTL (Af3-Af8) based on offspring from the cross-resistant line 29H × susceptible lineVf136.
Their QTL Af3 was located on chromosome III, just as QTL (Af1) mentioned above. Later, Díaz-Ruiz et al., (2009) confirmed, based on informative, common markers that QTL Af1 is located in the same chromosomal region as Af3. Furthermore, Díaz-Ruiz et al., (2009) confirmed the two QTL Af1 and Af2 in a recombinant inbred line (RIL) population derived from the material previously used by Roman et al., (2003). Both QTL together explained 24% of the phenotypic variation for disease severity on leaves and 16% of the phenotypic variation for disease severity on stem. Atienza et al., (2016) recently corroborated the QTL Af1 in a RIL population developed from the Avila et al., (2004) cross and judged it as identical to Af3. While Avila et al., (2004) assessed resistances after artificial inoculation in growth chamber, Atienza et al., (2016) tested in greenhouse and field. They re-identified Af2 on chromosome II and considered it the same as reported previously (Díaz-Ruiz et al., 2009;Kaur et al., 2014;Roman et al., 2003). Kaur et al., (2014) employed SNP markers and reported four Ascochyta QTL in a RIL population derived from crossing Icarus (susceptible) × Ascot (resistant) and tested under controlled conditions. Both belong to Australian-bred germplasm, seemingly different from the genotypes used in previous, Spanish studies. Their QTL-3 and the prior reported QTL Af2 (Díaz-Ruiz et al., 2009;Roman et al., 2003) were both located on chromosome II; however, lack of common markers prevents more definite conclusions (Avila et al., 2004;Díaz-Ruiz et al., 2009;Roman et al., 2003). Ascochyta blight resistance QTL-2 and QTL-4, detected first by Kaur et al., (2014), were confirmed in a biparental (Nurah × Farah) RIL population by Sudheesh et al., (2019), on chromosome I and VI, respectively. Since both parents of that population, Nurah and Farah, were resistant to strain1, that RIL population specifically segregated for resistance to Ascochyta blight strain 2.
Breeding for improved resistance to Ascochyta blight is thought to significantly support winter faba beans in Central Europe, in Germany and neighbouring countries. In times of global warming, winter beans may offer advantages over spring beans and increase the diversity of crop types to choose from. The Göttingen Winter Bean Population is a promising and highly relevant, diverse germplasm pool for this objective; it is used for breeding and research . In the present study, a panel of 188 winter faba bean lines, derived from that germplasm pool, is tested for Ascochyta blight resistance under controlled conditions and artificial inoculation. Phenotypic data on symptom expression are used for genome-wide association study to identify QTL for Ascochyta blight resistance in faba beans. Literature is thoroughly studied to exploit existing QTL data.
The study aimed to: 1. exploit existing data on Ascochyta blight resistance QTL from earlier studies in Mediterranean types and spring types of faba bean and 2. identify new markers associated with Ascochyta resistance in the Göttingen Winter Bean Population.

| Genetic material
The plant material consisted of N = 188 homozygous lines (association set; A-set) of faba bean (Vicia faba L.). These lines were bred via single-seed descent (SSD) without selection from the Göttingen Winter Bean Population (Link & Arbaoui, 2006) to generation F > 9.

| DNA markers and GWAS
Genotyping of the A-set lines as well as the 11 founder lines yielded a total of 2058 polymorphic markers, including 229 SNPs. Judged from the SNPs, the average degree of homozygosity was 98.8%, very high as expected. A number of 1,451 of these markers were mapped to 1,159 loci by Welna (2014). The 12 linkage groups of this author could be unambiguously assigned (Welna, 2014;pages XXVII-XXXV;cf. Ali, 2015) to the six Vicia faba chromosomes following the notation of Webb et al., (2016). After deleting markers with allele frequencies (MAF) ≤5%, a total of 1,355 markers were available for GWAS (1,147 AFLPs, 208 SNPs). The AFLP bands were scored as present/absent; the very few expected cases of wrongly scoring a heterozygous ALFP locus as homozygous dominant were tolerated. All AFLP markers and the majority of SNPs had been employed by Ali et al., (2016) for the same set of lines. All SNPs were chosen from the SNPs used by Webb et al., (2016). The SNPs contained 19 random SNPs and the so-called guide SNP marker set; these were not included in Ali et al., (2016). The 1,355 markers thus dissociate into two sets: R-set (randomly chosen markers set) and Gset (guide markers set). The R-set enclosed all AFLPs and the SNPs as taken at the onset of this research from Ali et al., (2016), plus the 19 additional, random SNPs. The G-set contained, after deleting three markers for their MAF ≤5%, 14 of initially 17 so-called 'guide' SNP markers. The new SNPs (19 plus 17) were analysed as reported by Ali et al., (2016).

| Development of the guide marker set (G-Set)
As all SNPs used here, the G-set markers can be found in Webb et al., (2016). The G-set markers were chosen during this study by exploiting prior literature about Ascochyta QTL (Table 1).
DNA markers linked to Ascochyta resistance QTL published in linkage maps Atienza et al., 2016;Satovic et al., 2013;more over Roman et al., 2003;Díaz-Ruiz et al., 2009, andAvila et al., 2004) were noted. The guide SNP markers (Table 1) Webb et al., (2016). A total of eight further guide SNP markers (four predicted to be near to Af2 and the other four predicted to be near to Af1) were thus defined from the mapped SNPs at Webb et al., (2016). All were predicted to be between 0.2 and 4.2 cM from Af1 or Af2, respectively.
The investigations of the maps of Roman et al., (2003), Díaz-Ruiz et al., (2009) and of Avila et al., (2004) did not reveal common markers, thus no additional guide SNP. Our association study is thus a study in two layers: a genome-wide association study with all 1,355 markers, and, included, a guided approach based on the 14 G-set markers.
Genome-wide association analyses were carried out using TASSEL version 3.0 (Bradbury et al., 2007). The mixed linear model (MLM) procedure of TASSEL was used with an optimum level of compression and re-estimation of the variance component estimates of each marker. A kinship matrix was employed, which was developed by using the average genetic similarity among the 11 founder lines as a threshold to define unrelatedness . A false discovery rate of 20% (FDR = 0.20) was used to test the statistical significance of marker-trait associations (Benjamini & Hochberg, 1995;Benjamini & Yekutieli, 2005).
Based on GWAS results and marker genotype, a marker score was calculated for the trait 'number of lesions per leaflet'. For this, for each inbred line, the sum of the effects of its markers with favourable allele present was calculated.

| Phenotyping
The experiments were conducted under semi-controlled conditions in showing typical lesions interspersed with pycnidia was air dried after sampling and subsequently incubated in humidity chambers for sporulation. Spores were picked from the ostiolum of a single pycnidium with the help of a sterile needle and transferred to V8-agar plates amended with 100 ppm streptomycin. In order to ensure to work with defined fungal genotypes, this procedure was once again repeated after one cycle of subcultivation. Two isolates named as number 50 and number 51 were used in the current study for being both, highly virulent yet differently responding to two rather susceptible and two rather resistant genotypes (Remer et al., 2016). The conidiospores were grown on V8 Agar media for the current analyses. Spore suspension was prepared with autoclaved tap water. The concentration of spore suspension was measured using a Fuchs Rosenthal hemocytometer under a microscope and further diluted to create the intended spore concentration (1 × 10⁶ conidia spores per ml). Spore suspension of isolates 50 and 51 was mixed as 1:1 ratio for inoculation. A fresh spore suspension was prepared for each inoculation event. Plants were inoculated one by one by spraying with spore suspension until the TA B L E 1 G-set of SNP (guide) markers as picked from the map by Webb et al., (2016)   where Y ij is the phenotypic value of a trait for inbred line i in replicate j, µ is the general mean, g i and r j are the main effects of genotypes and replications, respectively; and gr ij is genotype × replication interaction of genotype i with replication j.
This procedure allowed the software to estimate substitutes for the 0.92% of missing data points based on all 11 remaining replicates rather than based only on the one remaining of the two replicates per experiment. Variance components for genotypes were estimated from means squares of the analyses of variance as: Repeatability h 2 was estimated from mean squares (MS) and expressed in per cent: Spearman's rank correlation coefficients were used to examine correlations between traits.  (Table 3). Correlations within stem traits (0.72** < r < 0.91**) and within leaf traits (0.52** < r < 0.92**) were higher than between leaf and stem traits (0.44** < r < 0.61**) ( Table 3). The highest correlation was found for 'number of lesions per leaflet' with 'area covered by the lesions per leaflet' (r = 0.92**). However, moderate correlations were observed for the presence of pycnidia and other traits on leaflets (0.52** < r < 0.66**), whereas 'presence of pycnidia on stem' was highly correlated with the other stem traits (0.72** < r < 0.84**).

| RE SULTS
Association analysis was performed for the 188 A-set lines using their means across lattice-adjusted data from the 12 replications. The average LD among the 1,355 markers employed for GWAS was r 2 = 0.0075 (cf Ali et al., 2016). Among the 12 markers that were significantly associated with traits, the averaged LD value was r 2 = 0.0067, ranging from 0.0000 < r 2 < 0.108. A total of 12 markers, including nine AFLP and three SNP markers, displayed a statistically significant association with six of the eight traits; four of these 12 markers were associated with two or more traits (  Table 3). This one, marker 8 (E40M59-281), and the SNP Mt1g014230-001 (marker 5, associated with Af1) were in much higher LD than any other pair of markers (r 2 = 0.108). The AFLP marker E40M59-281 (marker 8) was furthermore associated with two leaf traits ('number of lesions per leaflet' and 'area covered by lesions per leaflet'; Table 4), which were as well highly correlated (r = 0.92**, Table 3). The explained phenotypic variance of marker 8 was, however, low; between 6.2% and 8.9% for its four associated traits. With the exception of marker 1 (E36M56-356), this range of explained variance is similar to that of other associated markers in the current study. For two stem-related traits (number of lesions and area covered by lesion), no significant marker-phenotype associations were detected.
The marker score for the trait 'number of lesions per leaflet' was correlated with the phenotypic result of this trait by r = 0.295**.  The Q-Q plot supports the notion that the database can be used for the applied analysis.

Trait
The average LD among all markers was very low (r 2 = 0.0075).
Given this very small LD, the available number of markers is likely a limiting factor. Indeed, two of eight traits were not associated with any of the marker, even though literature had been exploited for socalled 'guide' SNP markers.
This GWAS could be carried out with phenotype data of, mostly, high repeatability. The highest values h 2 were found for 'area covered by lesions per leaflet' (h 2 = 86.66%) and 'number of lesions per leaflet' (h 2 = 86.57%) (Table 2). Accordingly, the highest number of marker associations was detected for these two traits. Three statistically TA B L E 4 Association analysis results for Ascochyta blight resistance-related traits. Minimum minor allele frequency 5%, false discovery rate 20%. Chromosome number and cM position are according to Webb et al., (2016) (Welna, 2014).
f The AFLP alleles 1 (band present) or 0 (band absent) or the SNP allele A was the ones associated with decreasing (i.e. favourable) effect on trait. g This G-set SNP marker is probably associated with known QTL Af1 based on inferred map position and cross-inspection of Satovic et al., (2013) and Webb et al., (2016). significant markers were found for 'area covered by lesion per leaflet' and 10 markers for 'number of lesions per leaflet' ( was shared by 'number of lesions per leaflet' and 'area covered by the lesion per leaflet', and explained 7 to 8% of phenotypic variance. The exploitation of literature was successful insofar as this 'guide' SNP marker was picked for being listed in Webb et al., (2016) as near to the inferred position of a QTL marker of Satovic et al., (2013), when projecting that marker's position from the latter to the former map; even though the SNP was more than 4cM distant from that projected position. With one in 14 guide markers being significantly associated with resistance, this proportion was higher than for the genome-wide markers employed here.
Previously, QTL for Ascochyta blight have been reported on chromosomes II, III and VI. The QTL on chromosome III (Af1) has been reported and validated several times (Atienza et al., 2016;Avila et al., 2004;Díaz-Ruiz et al., 2009;Roman et al., 2003;Satovic et al., 2013). Kaur et al., (2014) assumed that their QTL3 in cross (Icarus ×Ascot) was identical with QTL Af2 on chromosome II; Af2 was reported by Díaz-Ruiz et al., (2009) and by Roman et al., (2003 Atienza et al., (2016) and by Avila et al., (2004 markedly lower than 85.6%. The sum of the effects of these 10 significant markers was 111.3 lesions per leaflet, which is higher than the maximum numbers (72.6 to 91.5) found in the most susceptible lines S_232, S060, S_168; these are indications of overestimation in the data. Furthermore, the correlation between the marker score for 'number of lesions per leaflet' and the phenotype itself was small, r = 0.295**, indicating that the marker score contains less information than what naïve interpretation suggests. The allele phases of two of the ten markers, marker 5 (presumably Af1) and marker 8, were associated (LD value of r 2 = 0.108), redundancy cannot fully be ruled out. With a false discovery rate of 20%, about two in 10 markers are not expected to be sustained as positives, and with the limited number of 188 lines, overestimation of effect sizes has to be anticipated (Josephs et al., 2017;Vales et al., 2005). Epistasis, as it is statistically presented as interaction, might be a further explanation for shrinking effects when joining markers (unless epistasis is specifically implemented in the statistical model; Göring et al., 2001).
Although the findings here are bound to the greenhouse conditions and the two fungal strains used, the currently most promising parents among the A-set lines to combine in a cross for breeding seem to be line S_150 and line S_162. This is because S_150 is, except for marker 4 (E44M58-177), homozygous for the resistanceassociated allele ('number of lesions per leaflet') at the other nine markers loci (Table 4); it is ranked as 26th best for the phenotypic trait value and on position 1 for marker score. The 25 lines that ranked phenotypically better than S_150 carried, at only three to eight marker loci, the resistance-associated allele; line S_150 was the highest-ranked line with nine. Line S_162 is, except for marker 5 (Vf-Mt1g014230-001) and marker 10 (E41M55-177), homozygous for the resistance-associated allele at the other eight loci. Line S_162 ranks third best for its marker score and first for its phenotypic value. Markers 4 and 5, although mapped on the same chromosome ( Marker-assisted introgression of resistance into elite genetic material is supported by conversion of AFLP-derived results into SNP-supported data. Currently, an Affymetrix 50K chip is under development and will be publicly available on short notice (O 'Sullivan, 2020;personal communication). Genotyping the A-set lines with that tool promises marked advance for applied breeding and for genetic analyses.

| CON CLUS IONS
Substantial and significant genetic variation for Ascochytaresistance traits was detected, and all eight assessed traits were seemingly genetically related. LD in this set of Göttingen Winter Bean lines was very low; the number of markers probably did not match such high genetic resolution. To reduce these limitations and to steer the focus towards Ascochyta blight resistance genes, a so-called guided marker approach was conducted in addition to the default genome-wide analyses. A total of 12 markers, including nine AFLP and three SNP markers displayed significant associations with six traits; nine of these markers probably stand for new resistance genes. Significant SNP markers were found at chromosomes III and VI in the descendants of the Göttingen Winter Bean Population. The guided approach was successful: one of 14 guide marker (Vf-Mt1g014230-001) was found significant and it is hypothesized that this SNP at chromosome III validates the previously reported QTL (Af1; chromosome III). The significant SNP found at chromosome VI should be validated in future studies.
Applied marker-assisted selection for Ascochyta-resistance relies strongly on the transfer of genetic results among different faba bean populations, depending on further saturation of QTL bearing chromosomal regions.

ACK N OWLED G EM ENTS
Technical assistance of Regina Martsch and Evelin Vorbeck is very thankfully acknowledged. First author very thankfully acknowledges stipend from the Government of Pakistan.

CO N FLI C T O F I NTE R E S T
Authors declare: no conflict of interest.

AUTH O R S CO NTR I B UTI O N S
Rabia Faradi conducted the experiments, gathered the data, con-

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.