Barley (Hordeum vulgare) is a major cereal crop with 145,6 million tons of grains produced on 48,9 million hectares in 2021 worldwide (FAOSTAT, https://www.fao.org/faostat/en/). It is primary used to feed livestock and to produce malt for the brewing industry. Various biotic and abiotic factors can reduce barley productivity and quality. Among these, barley “leaf scald” or “leaf blotch” is one of the major diseases in barley, particularly under wet and cool temperature climates. This disease was first reported as Marsonia secalis on rye in 1897 in the Netherlands and the same year in Germany on rye and barley (Caldwell 1937). It was then reclassified in the Rhynchosporium genus as Rhynchosporium graminicola based on morphological observations of typical beak-shaped one-septate conidia (Heinsen 1901). It was later renamed Rhynchosporium secalis due to its ability to infect barley, rye, triticale, and other grasses (Davis 1919). Phylogenetic analysis eventually led to reclassification into three different species based on host specificity and the one infecting barley and Bromus diandrus was named Rhynchosporium commune (Zaffarano et al. 2008; Braun 2016; Crous et al. 2021). Recent studies have shown that this pathogen originated from Northern Europe between 2,500 and 5,000 BP (Avrova and Knogge 2012). This disease is reported to cause up to 45% decrease in grain yield and reduces grain quality (Avrova and Knogge 2012).

Primary inoculum of R. commune originates from infected seeds (Skoropad 1959), from previous crop residues or from volunteer plants (Caldwell 1937). The spread of the disease from airborne spores is probably very limited (Fountaine et al. 2010). Conidia from infected leaves are then spread through splash-dispersal to upper parts of the plant, thus rainfall events during the spring are major factors of dissemination in a field. Conidia germinate in a temperature range of 4 to 28°C with an optimal temperature in the range of 18 to 21°C and under favourable moisture conditions (Caldwell 1937). This disease causes scalds mostly on leaf blades and to a lesser extend to leaf sheaths and ears starting by pale grey oval lesions often located near the leaf axil. Later the margin of the lesion becomes dark brown while the centre remains pale grey. Lesions merge leading to leaf chlorosis and eventually leaf death thus reducing photosynthesis, increasing plant respiration, and leading to reduced grain yield and quality (Avrova and Knogge 2012). Severe infections can lead to complete defoliation of the plant (Caldwell 1937).

Different management practices can be used to reduce the impact of barley leaf scald. Some agronomic practices can reduce the inoculum by burying previous barley plant residues or applying crop rotation. The use of fungicide in seed treatment or during vegetation can reduce the severity of barley leaf scald. However, fungicide resistant R. commune isolates have emerged and lead to reduced efficiency of these chemical products. The use of resistant cultivars remains one of the best options to control barley leaf scald since it can be very efficient while reducing the use of fungicides. However, breeding for resistant cultivars requires identifying sources of resistance in the available genetic material and precisely map the loci involved in resistance using genetic markers to facilitate and accelerate the selection of resistant cultivars. Given that splash-dispersal is a major spreading factor of this disease, leaf scald sensitivity is generally negatively correlated to plant height with tall genotypes being able to partly escape infection (Looseley et al. 2012; Walters et al. 2014; Looseley et al. 2015; Griffe 2017; Looseley et al. 2018). Earliness can also be considered as an escape trait as genotypes with a higher rate of development would reduce the duration of proximity of newly emerging leaves to sporulating lesions thus reducing epidemics (Royle 1994; Walters et al. 2014). It is thus important to consider these two traits to avoid cofounding effects when looking for genetic marker associated to leaf scald sensitivity.

One of the first avirulence genes that has been cloned was NIP1 (Necrose Inducing virulence Protein 1 or AvrRrs1) in R. commune (Rohe et al. 1995). Two other effectors inducing necrosis (NIP2 and NIP3) were later identified along with other putative effectors (Penselin et al. 2016). NIP1, NIP2, NIP3 code for effector proteins that contribute quantitatively to the virulence of R. commune depending on the host cultivar (Kirsten et al. 2012). This means that varietal resistance will depend upon the combination of Avr genes present in the pathogen populations. It is thus important to characterize the genetic diversity of pathogen populations to create resistant varieties to the most virulent strains. So far, this work has never been caried out for pathogen populations present in France.

Different resistance genes (Rrs) coming from cultivated (H. vulgare) and wild barley (H. vulgare ssp. spontaneum and H. bulbosum) have been identified. Rrs1 was the first identified resistance locus by bulk segregant analysis (Barua et al. 1993) and then confirmed using genetic mapping and genome wide association studies on chromosome 3H (Thomas et al. 1995; Graner and Tekauz 1996; Williams et al. 2001; Grønnerød et al. 2002; Hofmann et al. 2013; Looseley et al. 2018; Büttner et al. 2020; Thauvin et al. 2022). Rrs1 acts by preventing the penetration and subcuticular growth of R. commune strains carrying the NIP1 effector (Lehnackers and Knogge 1990; Thirugnanasambandam et al. 2011). Eleven Rrs1 alleles or tightly linked genes have been identified at this locus (Bjørnstad et al. 2002). Among these, Rrs1_Rh4 allele was originally found in Spanish and later in Syrian and Jordan material (Hofmann et al. 2013; Hofmann 2015; Looseley et al. 2020). Genetic markers partly linked to Rrs1_Rh4 were proposed after narrowing down the interval to 2.1cM using two doubled haploid populations (Hofmann et al. 2013) which is equivalent to 1.9Mbp (Looseley et al. 2018) in the Morex reference genome version 1 (Mascher et al. 2017). Since then, the size of the genomic interval carrying Rrs1_Rh4 was furthered narrowed down to 0.8 Mbp and diagnostic markers have been proposed (Looseley et al. 2020). Rrs2 was identified on chromosome 7HS (Schweizer et al. 1995) and later fine mapped to develop diagnostic markers for this locus (Hanemann et al. 2009). Candidate genes of the PECTN ESTERASE INHIBTOR FAMILY were suggested as responsible for this locus (Hanemann et al. 2009) but none of the tested gene could be identified as responsible for Rrs2 so far (Marzin et al. 2016). Other known Rrs genes coming from cultivated barley include Rrs3 on 4H (Bjørnstad et al. 2002), Rrs4 on 3H (Patil et al. 2003), Rrs15 on 7H (Genger et al. 2005), Rrs17 on 2H (Zhan et al. 2008; Wagner et al. 2008) and Rrs18 on 6HS (Coulter et al. 2019). Rrs12, Rrs13, Rrs14 and Rrs15 all came from H. vulgare ssp. spontaneum. Rrs12 was mapped on chromosome 7H (Abbott et al. 1991; Genger et al. 2003b), Rrs13 was located in a 22.5 cM interval on chromosome 6H (Abbott et al. 1991, 1995; Genger et al. 2003b), Rrs14 was mapped on chromosome 1H (Garvin et al. 1997, 2000) and Rrs15 was located on chromosome 7HL near the centromere (Genger et al. 2005). Rrs16 originally from H. bulbosum was introgressed into cultivar Emir through backcrossing and was mapped on chromosome 4HS (Pickering et al. 2006). Genome-wide association studies (GWAS) have shown additional QTL for leaf scald sensitivity (Griffe 2017; Looseley et al. 2018; Daba et al. 2019; Büttner et al. 2020; Hautsalo et al. 2021; Thauvin et al. 2022). Tables summarizing physical positions and flanking markers for the different Rrs genes and QTL have been published (Griffe 2017; Looseley et al. 2018; Zhang et al. 2020). The different loci identified so far confer a quantitative resistance phenotype. Thus, developing genetic markers closely linked to the genes behind these QTL or using genomic prediction models could help stacking these favourable genetic factors in elite cultivars thus improving the level of resistance and its durability (Brown et al. 1996; Genger et al. 2003b, a).

The objectives of this study were to (i) characterize the genetic diversity of R. commune isolates present in France, (ii) assess the sensitivity of a winter barley panel to leaf scald in the field and in controlled conditions, (iii) identify QTL and assess the accuracy of a genomic prediction model for leaf scald resistance.

Sampling and characterization of R. commune isolates

Ninety-nine samples of leaves showing apparent symptoms were collected in 2019 and 2020 on 40 cultivars in 43 locations in France. Pieces of leaves with necrosis were cleaned with Ethanol 96° for 20 seconds and bleach 0.5% for 1 minute and then dried using blotting paper. Three to four pieces were placed on PDA medium (Potato Dextrose Agar) in Petri dishes. They were incubated in growth chamber during 15 to 18 days in the dark at 17°C and under 60% humidity. Pieces of mycelium were then sampled and spread out in new Petri dishes containing PDA. After one week of incubation in the same conditions, one germinated spore or a piece of mycelium was sampled and spread out again in Petri dishes with PDA. Finally, plugs of agar with mycelium were sampled and stored at -80°C in 2mL cryotubes containing glycerol 25%. The genetic diversity of R. commune isolates for NIP1, NIP2 and NIP3 genes was assessed after PCR amplification (Stefansson et al. 2014) and sequencing of amplicons by the Eurofins company. Sequences were then analysed using the Geneious software to identify polymorphic sites and the possible impact of these mutations on protein function. The nucleotidic diversity was calculated as:

With d_ij the number of nucleotidic differences between sequence i and j, L the length of the sequence, n the number of sequences and the number of pairs of sequences. The haplotypic diversity was calculated as:

With n the number of haplotypes, k the number of unique haplotypes, p_i the frequency of haplotype i.

Genotypic sensitivity to leaf scald was assessed visually for 202 to 289 accessions in 14 different field experiments performed in 11 locations during 5 years in France (Table 1). Experiments were arranged as a randomized complete block design with two replicates except for trials LIM17, RAG18 and SYN18 where only a single plot was present for each accession. Plots consisted in three 1m rows with an interrow distance of 20 to 25cm and a seed density of ca. 250 seeds per square meter. Rhynchosporium symptoms observed in the trials resulted from natural inoculum present in the different fields and therefore pathogen population could be different from one field to another. Scoring was performed visually using a scale ranging from 1 (no symptoms) to 9 (leaves fully covered by necrosis) on the whole canopy (Thauvin et al. 2022). One to three dates of scoring were performed during the stem elongation period until heading. Only the date with the highest median score and largest variance was considered. Heading date (growth stage 55 (Zadoks et al. 1974) expressed in degree days since sowing base 0°C) and plant height at heading (from the base of the plant to the top of the spike, awns excluded) were also recorded in most of the trials as they could act as cofounding factors on leaf scald sensitivity (Table 1). Regarding trials where plant height and/or heading date were not recorded, average values over the whole trial network were used as covariates. In each trial, best linear unbiased estimates (BLUEs) of each trait were calculated using a mixed linear model with genotype as fixed and replicate as random effects using the lme4 package (Bates et al. 2015) in R (R Core Team 2016). Then, BLUEs of leaf scald sensitivity were corrected for the effect of escape traits using the method proposed by Beuningen and Kohli (1990) and also used in the particular case of leaf scald in Thauvin et al. (2022). First, scores were normalized by calculating the relative coefficient of infection (RCI):

With the relative coefficient of infection of genotype i in trial j. the sensitivity score of genotype i in trial j and the maximum sensitivity score in trial j. Then, residuals of the multiple linear regression explaining by plant height and heading date were calculated for each genotype in each trial:

With the relative coefficient of infection of genotype i in trial j, heading date of genotype i in trial j and plant height of genotype i in trial j and the residual of the regression hereafter called DRIHH (deviation from the regression of infection on heading and height) to follow the terminology proposed by Beuningen and Kohli (1990). DRIHH thus represents the residual variation of (i.e. leaf scald sensitivity) after considering the effect of heading date and plant height. Positive values represent higher leaf scald sensitivity than expected given plant height and heading date of a given genotype.

Broad sense heritability was calculated as:

With the genotype variance, the residual variance of a mixed effect model including genotype (random) and trial (fixed) effects and the number of trials.

Leaf scald sensitivity of the panel was assessed for three isolates of Rhynchosporium commune (RHY19001, RHY19012, RHY19026). For each isolate, different trials had to be carried out as only 25 genotypes could be tested at the same time in the growth chamber. Cultivars Etincel and Isocel were present in all the trials and were used as controls to adjust differences between trials. A solution containing spores at a concentration of 1.10⁶ spores.mL^− 1 was prepared for each isolate. Five seeds of each accession were sown in 10cm squared pots. Pots were watered and maintained in growth chamber under 16h photoperiod, 22°C day, 18°C night, 80% humidity day, 60% humidity night. After 12 to 14 days, an area of 7cm long was defined on the last leaf and spores were dropped after six passages of a paint brush on this area. Pots were then covered with a black plastic bag and maintained as such in growth chamber for 72 hours. The percentage of necrosis in the delimited area was scored visually after 21 days post inoculation on the five plants and an average score was calculated for each variety.

A mixed model including genotype (fixed) and trial (random) effects was used to correct genotypic sensitivity scores obtained in the different trials carried out with the same isolate. This model was adjusted using a Bayesian method with the R package brms and medians of the posterior distribution were calculated for each genotype. Adjustments were obtained from 5 chains with 2,000 iterations each and using a uniform prior distribution with values ranging from 0 to 1 while the other brms parameters were set as default.

The panel was genotyped using the Illumina Infinium iSelect 50K SNP array (Bayer et al. 2017) at the James Hutton Institute. Additionally, 200 accessions were genotyped using 440 KASP markers. Raw genotyping data included 43 799 markers with an average missing data rate of less than 1% (41 780 markers had no missing data at all). Markers were filtered to keep those showing less than 10% of heterozygous genotypes and a minor allele frequency above 5%. This led to a total of 31 955 SNP markers available for GWAS (4 000 to 8 000 markers per chromosome). Missing genotype data were imputed using Beagle (Browning 2008). Markers were mapped on the Morex v2 reference genome (Monat et al. 2019). Population structure was assessed using the STRUCTURE software (Pritchard et al. 2000). Ten runs for K varying from 2 to 15 sub-populations were performed and the number of sub-populations was determined using the method proposed by Evanno et al. (2005).

GWAS was performed for DRIHH in each field trial and for the percentage of leaf necrosis for each isolate tested in controlled conditions. The GWAS model was a mixed model including SNP, genetic structure and kinship effects:

With the trait value of genotype i in trial or isolate j, the overall mean in trial or isolate j, the fixed effect of marker m in trial or isolate j, the genotype of cultivar i at marker m, coordinates of genotype i on principal component k (two principal components were used), is the random effect of genotype i with covariance derived from the kinship matrix and are independent and identically distributed residuals following normal distribution. The kinship matrix was calculated using the rrBLUP package in R. This model was adjusted using the R package MM4LMM for each marker (Laporte et al. 2022). A significance threshold of -log10(p) = 5.09 was calculated using the method proposed by Gao et al. (2010) and Gao (2011) and a threshold of 5 was retained. The additive effect of markers was calculated as follow:

With the average leaf scald sensitivity of individuals homozygous for the A allele and the average leaf scald sensitivity over all the individuals.

A GWAS meta-analysis was performed separately for leaf scald sensitivity measured in the field and in controlled conditions using the metaGE package in R (Walsche et al. 2025). This method uses p-values and marker effects and allows identifying markers significant across different environments where the same panel of genotype was tested. This can lead to the identification of additional QTL compared to single environment analysis. The “random effect” model was fitted with metaGE thus accounting for possible heterogeneity of the QTL effects across trials/isolates. The Benjamini-Hochberg procedure was applied to identify significantly associated markers with p-values ≤ 0.05. The significance threshold was determined as -log10 of the max p-value among significant markers.

QTL boundaries were defined based on linkage disequilibrium (LD) following the method described in Cormier et al. (2014). LD was calculated after correcting for population structure and individual relatedness using package LDcorSV in R (Mangin et al. 2012). First a “critical LD” threshold was calculated as the 95th percentile of the unlinked-r2 assessed on 1,000,000 randomly chosen pairs of unlinked loci (mapped on different chromosomes) which were square root transformed to approximate a normally distributed random variable. The GWAS results for DRIHH and for the percentage of leaf necrosis for each isolate tested in controlled conditions were combined. Then, significantly associated SNPs were grouped into clusters based on average distance using a cutoff of 1-“critical LD”. For each cluster, LD was calculated for every pair of markers in a region starting from the minimum to the maximum position +/- 1Gbp of the considered cluster. The LD decay was then modelled using an exponential function, the physical distance corresponding to LD equal to 0.2 was calculated and this value was used to extend the QTL boundaries following recommendation in Alqudah et al. (2020). QTLs were named by concatenating the name of the chromosome and the position of the most significant marker (in Gbp) on the Morex v2 reference genome (Monat et al. 2019). Analysis of QTL co-location between trials/isolates and known QTLs or Rrs genes was performed. Morex v2 high confidence predicted genes present in the interval of each QTL were extracted. Genes with predicted functions related to disease resistance (“Disease resistance protein”, “Leucine Rich Repeat receptor”, “Receptor kinase”) were considered as possible candidates for the QTLs.

BLUEs for leaf scald sensitivity, heading date and plant height across trials were calculated using a mixed linear model with genotype as fixed and trial as random effects using the lmer R package. DRIHH was then calculated using the across trial average leaf scald sensitivity, heading date and plant height. A GBLUP model was calibrated for DRIHH using the BGLR package in R (Pérez and de los Campos 2014). The model was cross-validated (CV) using a 10-folds CV scheme to assess its ability to predict leaf scald sensitivity for untested genotypes.

Genetic diversity of NIP genes in R. commune isolates present in France

A total of 103 single spore isolates of R. commune were obtained from the 99 samples collected in France in 2019 and 2020. The genetic diversity of 81 isolates of R. commune for NIP1, NIP2 and 71 isolates for NIP3 was assessed (Table 2). The length of the coding region of NIP1 was estimated to 249bp (Table 2). NIP1 was absent of five isolates and seven showed truncated NIP1 or premature stop codons (Table 2). The remaining 69 isolates showed 15 different haplotypes (haplotypic diversity equal to 0.761 and nucleotidic diversity equal to 0.01035, Table 2) among which two appeared in 39 and 30% of the isolates tested and nine were present in only one isolate. The NIP2 sequence was 330 bp long with no intron (Table 2). No isolate showing complete deletion of NIP2 was identified. Only two SNPs corresponding to non-synonymous mutations were detected (Table 2). Consequently, one haplotype was present in 93% of the tested isolates, a second one in 6% and a third one in one isolate. The NIP3 sequence was 409 bp long and showed two exons separated by one intron of 61 bp leading to a coding region of 348 bp. Over the 71 isolates sequenced for NIP3, only one SNP with a non-synonymous mutation was found in 19 isolates (Table 2).

The analysis of population structure showed stratification into two groups corresponding to two-rows and six-rows barley cultivars (ESM 2). A total of 122 significant markers located on all barley chromosomes were identified by single environment analysis and/or GWAS meta-analysis. Single environment GWAS analysis identified 88 significant markers for leaf scald sensitivity in the field (maximum -log10(p) = 11.8) and 24 significant markers in controlled conditions (maximum -log10(p) = 7.5) (ESM 3). Three and seven additional markers were identified by GWAS meta-analysis for leaf scald sensitivity in the field and in controlled conditions, respectively. Fourteen markers on chromosome 1H, one on 2H and three on 7H were significant in two to three trials (ESM 3). Four markers on chromosome 3H, three on 5H and one on 7H were significant for two isolates tested in controlled conditions (ESM 3).

All the significant markers were grouped into 19 QTLs based on linkage disequilibrium (Table 3, Fig. 4). The additive effects of the QTLs for leaf scald sensitivity scored visually in the field ranged from 0.54 to 1.69 thus showing minor impact of individual QTLs (Table 3). On chromosome 1H, Qrrs-1H-8 located between 8,211,982 and 9,735,919 bp in the Rrs14 region was significant after GWAS meta-analysis for leaf scald sensitivity in controlled conditions. Qrrs-1H-483 significant in four trials (ASU21, LIM21, RAG18 and RAG23) was located between 480,297,677 and 487,464,881 bp with max -log₁₀(p) equal to 11.81 (Table 3). Markers for this QTL were in the FT3 (Ppd-H2) region although they were not significant for heading date in these trials (Fig. 4). Four QTLs were found on chromosome 2H with max -log₁₀(p) ranging from 5.02 to 6.70 (Table 3). Qrrs-2H-14 was located between 14,337,143 and 14,399,788 bp within the Rrs17 region (Zhan et al. 2008; Wagner et al. 2008; Griffe 2017) and was significant in FLO21 (Fig. 4). Qrrs-2H-611 located between 610,909,011 and 611,155,121 bp and significant in LIM17 was ca. 1.5Gbp away from Qsc-2H.5 (Daba et al. 2019). Qrrs-2H-638 was between 638,097,711 and 638,330,069 bp ca. 2Gbp away from QTLRS7a (Sayed et al. 2004) and significant in LIM17. Qrrs-2H-674 was located between 669,984,878 and 674,705,254 bp ca. 800Kbp from QTLG2H (Gawenda et al. 2015) and significant in three trials (ARV22, SYN22, SEC22). Three QTLs were found on chromosome 3H. Qrrs-3H-6 located between 5,538,612 and 5,780,554 bp was identified after GWAS meta-analysis for leaf scald sensitivity in the field (Fig. 4). Qrrs-3H-15 was located between 13,909,134 and 154,409,718 bp about 400Kbp away from QTLSR3H.1 (Looseley et al. 2015) and was significant in two trials (RAG18 and SEC23) with max -log₁₀(p) = 5.79 (Table 3). Qrrs-3H-447 was between 444,759,929 and 455,880,328 bp in the Rrs1 region (Genger et al. 2003a; Looseley et al. 2020) and significant for two isolates (RHY19001 and RHY19012) with max -log₁₀(p) = 7.53 (Fig. 4). On chromosome 4H, Qrrs-4H-621 was located between 618,282,891 and 620,684,869 bp and significant in two trials (SEC22 and RAG23) with max -log₁₀(p) equal to 5.55 (Table 3). Two QTLs were found on chromosome 5H with max -log₁₀(p) ranging from 5.43 to 6.37 (Table 3). Qrrs-5H-570 was located between 569,294,646 and 569,943,458 bp and was significant in LEM23. Qrrs-5H-576 was between 574,281,246 and 576,313,778 bp within the Qsc_5H_1 interval (Hautsalo et al. 2021) and significant for two isolates. Three QTLs were found on chromosome 6H (Table 3). Qrrs-6H-4 was located between 3,644,390 and 3,869,372 bp approximately 1Gbp away from Qryn6 (Jensen et al. 2008) and was significant in SEC23. Qrrs-6H-482 was located between 471,277,250 and 485,578,495 bp and was significant in ARV22. Qrrs-6H-569 was between 568,911,666 and 569,133,702 bp and significant after GWAS meta-analysis (Fig. 4). Four QTLs were found on chromosome 7H with max -log₁₀(p) ranging from 5.91 to 7.56 (Table 3). Qrrs-7H-4 located between 287,172 and 5,201,201 bp within QTLIS7H.2 (Bjørnstad et al. 2004) and 1Kbp away from Rrs2 (Hanemann et al. 2009) was significant for two isolates (Fig. 4). Qrrs-7H-11 located between 10,796,027 and 10,932,119 bp about 3 Gbp away from Rrs12 (Fig. 4) was significant in two trials (RAG23 and SYN22). Qrrs-7H-595 was located between 589,674,278 and 597,377,376 bp ca. 1Gbp from Qsc-7H.4 (Daba et al. 2019) and was significant in three trials (RAG23, SEC22 and SYN22). Finally, Qrrs-7H-606 was located between 602,006,686 and 607,041,406 bp ca. 100Kbp from QI9 (Looseley et al. 2018) and was significant in three trials (SEC23, SEC22, RAG23).

A total of 82 possible candidate genes with functions associated to plant pathogen interactions were identified in the interval of ten different QTLs (ESM 4). HORVU.MOREX.r2.1HG0062420 a “receptor-like protein kinase” was present in the interval of Qrrs-1H-483. Six genes coding for “receptor kinase” proteins and five genes coding for “nbs-lrr disease resistance protein” were found in Qrrs-2H-674. In Qrrs-3H-15, 13 genes coding for “receptor kinase” and one “wall-associated kinase” were identified. Five “receptor-like kinase” were found in the interval of Qrrs-3H-447 in the Rrs1 region. Three genes coding for “leucine-rich repeat receptor protein kinase family” proteins, one “receptor-like kinase” and one “l-type lectin-domain containing receptor kinase viii.2” were found in Qrrs-4H-621. Four genes coding for “leucine-rich repeat receptor protein kinase family” proteins and two “receptor lectin kinase” were found in Qrrs-5H-576. One “protein enhanced disease resistance 2-like” was present in the interval of Qrrs-6H-4. One gene coding for a “lectin receptor kinase” protein was found in Qrrs-6H-482. Many genes were found in the Qrrs-7H-4 region with 24 genes coding for “disease resistance” proteins, eight for “receptor kinase”. Finally, Qrrs-7H-606 contained one gene coding for a “leucine-rich repeat receptor-like protein kinase” and five genes coding for “receptor kinase” proteins.

Diversity of NIP genes in R. commune isolates present in France

Characterization of NIP1, NIP2 and NIP3 diversity in R. commune isolates collected in France was carried out in this study by sequencing these genes in 81 (NIP1, NIP2) and 71 isolates (NIP3). These isolates were collected from samples of different winter and spring barley cultivars in different locations in France in 2019 and 2020. Different polymorphisms of NIP1 were identified including total deletion of the gene and non-synonymous SNPs leading to a total of 22 protein haplotypes including seven non-functional proteins. This showed the large genetic diversity of NIP1 in R. commune isolates in France. On the contrary, NIP2 and NIP3 showed much lower degree of genetic diversity in the isolates tested with only two and one SNPs detected, respectively.

NIP1 is both an effector and an elicitor which means that it can facilitate infection while trigger host resistance. Deletion or mutation of an avirulence gene can allow pathogens to escape their recognition by the corresponding host resistance gene. Thus, the strong genetic diversity of NIP1 probably indicates a strong selection pressure due to Rrs1 which is present in the panel of accessions tested (Fig. 4). Indeed, 28% of the accessions had the resistance allele at Qrrs-3H-447 (Fig. 6). The low diversity in NIP2 could be due to a lower selection pressure of Rrs2 which is less present in cultivated barley in France. Indeed, genotyping of the panel using a KASP marker derived from the diagnostic markers of Rrs2 identified by Hanemann et al. (2009) showed only 6 accessions with Rrs2 (data not shown). In addition, the lower polymorphism in NIP2 is in line with other data in the literature showing a lower genetic diversity of this gene compared to NIP1 (Stefansson et al., 2014).

This study allowed collecting 103 isolates of R. commune from different locations and cultivars in France in 2019 and 2020 and characterizing the genetic diversity of NIP genes for 81 isolates. High genetic diversity was observed for the NIP1 gene possibly indicating a strong selection pressure due to the presence of Rrs1 in the cultivars registered in France. A genome-wide association study carried out for a panel of 289 winter barley accessions tested in 14 field trials and for three isolates in controlled conditions allowed identifying 122 genetic markers which were clustered into 19 QTLs with minor effects and spread over all the chromosomes in the barley genome. Most of these regions co-located with already known loci for leaf scald sensitivity such as Rrs1, Rrs2, Rrs12, Rrs14, Rrs17 while a few others could represent new sources of resistance with some being at low frequency in the panel tested. The average genotype leaf scald sensitivity was highly correlated to the number of QTL with resistance allele indicating additive effects of the different QTLs. The accuracy of a genomic prediction model for average genotype leaf scald sensitivity showed that genomic selection could be an interesting breeding strategy for this trait.