LLS scoring system
Percentage of leaf area with P. brassicae sporulation observed on infected leaves was the most consistent measurement of disease severity across all accessions (Karandeni Dewage et al. 2021) and was therefore used as the basis for the scoring system. Disease severity was scored on a scale of 1 to 6, with a score of 1 for no sporulation and 6 for the most sporulation (Fig. 1A). The image of part of an infected cv. Tapidor leaf (Fig. 1A, score 2) shows why P. brassicae is referred as to C. concentricum in its imperfect stage; this pathogen produces concentric rings of acervuli on its hosts. Patchy sporulation was observed on leaf laminas of ‘couve-nabiça’ and cv. Capitol with scores of 3 to 4, respectively (Fig. 1A). The entire leaf laminas of cv. Musette and cv. Daichousen were covered with acervuli and scored 5 and 6, respectively. Whilst the resistance response black necrotic flecking was sometimes observed during the screening process, it was not present on all of the accessions tested, and so this phenotype was not considered suitable for scoring QDR.
Variation in sporulation of P. brassicae on leaves of diverse B. napus accessions
A diversity set of 195 accessions was tested in glasshouse experiments for the amount of pathogen sporulation after spray inoculation of B. napus seedlings with local P. brassicae populations. The distribution of disease scores showed wide variation in P. brassicae sporulation among diverse B. napus accessions with an approximately normal distribution of this trait (Fig. 1B).
The glasshouse screen consisted of 10 independent experiments, each with 23 or 24 B. napus accessions with four reference cultivars per experiment (Supporting Information Table S2). The first seven experiments were done at the Bayfordbury campus of the University of Hertfordshire and the last three experiments at Rothamsted Research. Instead of cv. Bristol, cv. Cabriolet was used as a susceptible reference cultivar for the experiments at Rothamsted Research. All other reference cultivars were identical between both sites. Irrespective of the location, there were significant differences between reference cultivars in disease score. No significant effects of experiment and experiment-by-cultivar interaction on disease score were observed (Supporting Information Notes S1). Combination of all 10 experiments and comparison of the shared reference cultivars Imola, Tapidor and Temple resulted in significant effects of both cultivar and experiment, but no significant experiment-by-cultivar interactions on disease score (Supporting Information Table S2). The disease scores of all three reference cultivars were less at Rothamsted Research than at Bayfordbury, with cv. Temple differing the most between these two environments (Supporting Information Figure S1). Restricting the analysis to cv. Imola and cv. Tapidor eliminated the significant effect of experiment. Differences in disease scores between Rothamsted Research and Bayfordbury could have resulted from different environmental conditions, pathogen inoculum and assessors at the two sites.
Non-linear mixed model analysis established that B. napus cv. Cabriolet and cv. Imola scored as the most susceptible and most resistant cultivars amongst the five reference cultivars tested, respectively (Supporting Information Table S2). Intermediate scores were observed for the other three cultivars Temple, Tapidor and Bristol. While cv. Cabriolet scored as the most susceptible cultivar amongst all 195 accessions tested, 60 accessions scored less sporulation than cv. Imola; half of these 60 accessions scored significantly less than cv. Imola including all 11 accessions that supported the least P. brassicae sporulation. These data clearly show that quantitative resistance present in 30 diverse accessions resulted in less P. brassicae sporulation than in cv. Imola, which has a major QDR locus against this pathogen.
GWA mapping of quantitative resistance against P. brassicae
SNP data for the RIPR genotype dataset from the resources page of York Knowledgebase (http://yorknowledgebase.info) were used for this study. Following analysis with STRUCTURE, calculation of ΔK divided the population into two clusters (Figure 2); cluster one mainly comprising of winter OSR and fodder types and cluster two comprising of other crop types (Supporting Information Table S3). It was shown that K = 2 is a common outcome when using the ΔK method (Janes et al. 2017). ΔK frequently identifies K = 2 as the top level of hierarchical structure, and further analysis is required to determine whether more subpopulations are present. Our analysis identified a further maximum in ΔK at K = 6. This divided the population into groups comprising the different crop types; cluster one - winter and fodder; cluster two – swede; cluster three – spring OSR; clusters four and five – Siberian kale types, and six semi-winter (Chinese) OSR (Supporting Information Table S3). Subsequent phylogenetic analysis showed a delineation of these crop types (Figure 2), with the different crop types forming clear subgroups within the tree. A number of accessions did not cluster with their given crop types. However, bar diagram outputs from STRUCTURE (Supporting Information Figure S3) showed a level of admixture within these accessions. Given the evidence for population substructure beyond K = 2, a structure of K = 6 was taken forward for use in association mapping.
GWA mapping using TASSEL identified a generalised linear modelling approach as an optimal fit for the phenotypic data (Supporting Information Figure S4). Eleven significant marker associations with LLS infection score were observed at P < 0.0001 (Table 1); however, none of these were significant at the FDR 0.05. This is not unexpected as resistance against P. brassicae is a highly quantitative trait, with no known R gene loci.
The allelic effects of identified GWA maxima were determined (Table 2). The distribution of alleles contributing to resistance was not determined by crop type or phylogenetic relationship. Due to sequence similarities, cross alignment of transcriptome reads occurs between homeologous loci in the A and C genomes. This means that allelic calls can be the same in each genome or carry alternate alleles in the A and C genomes, referred to as a hemi-SNP, resulting in an ambiguity call during SNP calling. Three loci, LLSC01, LLSA09a and LLSC02, showed the strongest resistance when present as hemi-SNPs, suggesting a resistance benefit linked to carrying an alternative allele at the homeologous locus. Loci LLSA01 and LLSC04 showed that the majority of OSR in the panel carried the resistant allele; therefore these may have already been selected for during breeding. Lines carrying alternate alleles at the homeologous position were also present, suggesting some breeding lines may not be optimised for these potential resistance loci. Loci LLSA02, LLSA07, LLSA09b and LLSC08b carried both A and C resistance alleles in a small number of lines, with most lines carrying alternate alleles at the homeologous loci. For locus LLSC08b, only one line carried susceptible alleles in both sub-genomes, resulting in an elevated disease score. For loci LLSC05 and LLSC08a, resistant alleles at the two homeologous loci were not present within winter-OSR (WOSR) or in the case of LLSC08a, within the panel tested. Hemi-SNPs associated with enhanced resistance indicate currently un-described cryptic loci that could be further optimised for resistance in breeding unused loci for resistance breeding against P. brassicae.
Although the 11 GWA markers detail the genetic variation most significantly associated with resistance against P. brassicae within this analysis, transcriptome sequencing does not provide all potential genetic variants present across the panel. Genetic polymorphisms close to the causal variation will be associated on the basis of genetic linkage or LD. In the case of LD, the GWA markers may not be indicative of the causal gene. Instead, causal genes will be more or less closely linked to the GWA marker. Each GWA marker was tested for LD against all other markers on a chromosome. Four of the 11 significantly associated markers, loci LLSA02, LLSA09, LLSC02, LLSC08a, exhibited LD with r2 > 0.15, thus defining the region where the causal gene is likely to be situated (Table 3, Figure 3).
Eight gene expression markers are associated with resistance or susceptibility to P. brassicae infection
The transcriptomes of the 195 accessions were used to analyse the disease scores as a function of the expression of all gene models using a general linear model with a false discovery rate of 0.001. A list of eight genes with P-values < 1.6 x 10−7 was selected for further analysis (Table 4). Linear regressions of the disease score versus the expression of each gene generated different slopes (Supporting Information Figure S2); the expression of seven genes was negatively correlated with the disease score. Thus, expression was greatest when the least sporulation was observed. These seven GEMs may therefore contribute to partial resistance against P. brassicae. Another gene, encoding an HXXXD-type acyl-transferase (Table 4), had the opposite expression pattern with an expression positively correlated with the disease score. This is therefore a candidate gene for susceptibility to P. brassicae.