QTL mapping and transcriptome analysis of Sclerotinia-resistance in the wild cabbage species Brassica oleracea var.

Oilseed rape (Brassica napus) is one of the most important oil-producing crops worldwide. The narrow gene pool of oilseed rape hampers its resistance breeding. Sclerotinia stem rot (SSR), caused by Sclerotinia sclerotiorum, is one of the most destructive diseases in many oilseed rape growing regions, worldwide. So far, no effective genetic source of resistance to S. sclerotiorum in B. napus germplasm is available, and yet knowledge of molecular plant-fungal interactions is limited. To identify new resistance source against SSR, we generated a segregating F 2 population for Sclerotinia resistance with 510 individuals from an interspecic cross between the resistant B. villosa (BRA1896) and a wild susceptible B. oleracea (BRA1909). Genetic mapping using a 15k Illumina Innium SNP-array resulted in a high-density genetic map that contains 1,118 markers and spans a total genetic length of 792.2 cM. QTL-analysis identied 7 QTLs for Sclerotinia-resistance and 5 QTLs for trichome-phenotype, which explain up to 16.85 % and 34.45 % of corresponding phenotypic variance, respectively. Although a partial co-localization of major QTLs for trichome-phenotype and Sclerotinia-resistance was given, no functional association between these two traits could be validated. In addition, comparative RNAseq analysis suggests that activation of JA- and ethylene-mediated responses plays a central role in the Sclerotinia-resistance, associated with a stronger plant immune response, depressed cell death and elevated phytoalexin biosynthesis in B. villosa. Our data demonstrate that the wild Brassica oleracea complex represents a novel and unique genetic source of Sclerotinia resistance for breeding resistant oilseed rape against SSR. via marker assisted selection (MAS) into the B. napus gene pool Comparative transcriptome analysis in the wild B. oleracea species linked this resistance to an early perception of the pathogen followed by a rapid release of reactive oxygen species (ROS) modulated by Ca 2+ signaling pathways and an early suppression of Sclerotinia virulence genes (Ding et al. 2019b). These studies highlight the B. oleracea gene pool as important source for introgression of improved resistance to Sclerotinia into the primary gene pool of B. napus. Herein, we present results from phenotypic and genetic analyses on a segregating F 2 -population from an intercross between the wild B. villosa, highly resistant to Sclerotinia, and the wild, susceptible B. oleracea var. oleracea. For the rst time, we report QTLs for Sclerotinia-resistance and trichome-phenotype in the genome of B. villosa. In addition, the comparison with previously identied QTLs in the wild B. incana (Mei et al. 2013, Mei et al. 2017) allows for evaluating the resistance mechanisms existing in different Brassica species. Our genetic studies are complemented by an integrated reference- and de novo-based comparative transcriptome proling 8 hours post inoculation (hpi), providing unique insights into the early defense response in the resistant B. villosa. We identied 58 B. villosa-specic signicantly upregulated defense-related genes that are absent in the reference genome of B. oleracea and demonstrate that the distinct activation of the signaling pathways by jasmonic acid (JA) and ethylene (ET) plays a pivotal role in Sclerotinia-resistance and is associated with a strong immune response, a negative regulation of cell death, and an elevated phytoalexin biosynthesis. A F 2 510 individuals interspecic B. villosa x B. into two populations with 252 258 F 2 -individuals, as Population A and B, and cultivated under greenhouse conditions. From Population A, 234 F 2 -plants were once evaluated for Sclerotinia-resistance via the detached leaf- and petiole-assay, from these 187 were selected for genotyping with the B. napus 15k Illumina Innium SNP-chip. From Population B, 258 F 2 -plants were evaluated once via the detached leaf- and petiole-assay. From these, 184 individuals were selected and genotyped. The phenotypic data from the genotyped Population B was re-evaluated with the leaf- and petiole-assay under greenhouse twice. After resistance 171 individuals in B were scored according to their trichome-phenotype. SyGreen Mix (PCR Biosystems Inc., Pennsylvania, USA). qPCR was performed on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, California, USA). Conditions for the reactions were as follows: 3 min at 95 °C; 40 cycles of 95 °C for 15 s, 60 °C for 15 s, and 72 °C for 20 s. Relative quantication was calculated in relation to Actin7 (Bo3g005290) according to Pfa (2001), which showed a stable expression in the RNAseq. Analysis of primer ecacy was determined by a standard curve of pooled cDNA from all samples for each gene. Primer and their predicted targets in the RNAseq are provided in the supplement (Supplementary Data S6). Statistical analysis was performed via a linear model using generalized least squares with the nlme-package (Pinheiro et al. 2020). Sequencing of selected templates validated targets in the wild B. oleracea species. The log2-transformed gene expression from the qPCR data at 8 hpi was compared to the RNAseq data. in the 1 st petiole-assay increased the LOD from 2.79 to 3.53 below the LOD-threshold of 3.56 (p = 0.053). Both peaks from the 1 st and 3 rd petiole-assay on C06 at 46.8 (35,415,831 bp) and 47.8 (35,610,420 bp) additive effect in Population B. The QTLs on C01 explained about 15.8 % (pQTLa) and 16.5 % (l2QTLb) of variance. The QTLs on C03 in in comparison to B. oleracea. In the susceptible host, we identied a 50-times more signicantly enriched response to SA (GO:0009751) and a three-times more signicantly enriched response to JA (GO:0009753) in B. oleracea. Both species showed a highly signicant (log10-FDR < -30) response to decreased oxygen levels (GO:0036293). The immune response (GO:0006955) was strongly addressed in both species, but 7.6 x 10 11 times more signicantly in the resistant B. villosa. Regulatory processes of the immune system (GO:0050776), a negative regulation of cell death (GO:0060548), and a positive regulation of defense response (GO:0031349), as well as a response to reactive oxygen species (ROS; GO:0000302) were specically enriched in B. villosa. The resistant BRA1896 showed overall a strong increase in the primary metabolism (e.g., GO:0015979; GO:0016168; GO:0047899; GO:0009538) and in the phytoalexin metabolic process (GO:0052314; GO:0046217; GO:0052317). The susceptible B. oleracea showed a specic enrichment of the glycosinolate metabolic process (GO:0019757) and the ABA-activated signaling pathway (GO:0009738). The GO enrichment analysis of the 854 genes with signicant differences in their log2-fold change expression revealed multifaceted transcriptomic strategies in the commonly enriched terms of both species due to the inoculation. For example, 53 genes associated with the hormone-mediated signaling pathway (GO:0009755) showed a signicant difference in their expression between the two species as result of the inoculation. Further differences were identied in the signal transduction (GO:0007165; 94 genes), in the response to decreased oxygen levels (GO:0036293; 28 genes), in the response to ABA (GO:0009737; 42 genes), and in the ET-activated the trichome-phenotype B. villosa of the TRY stronger in the resistant B. villosa. identied 17 signicantly differentially ET response transcription factors (ERF) B. villosa and B. oleracea the Sclerotinia-inoculation. of these ERFs were positively induced in B. villosa. different expression proles and the increase of expression of the ethylene receptor gene ETR2 in B. villosa at 16 hpi the assumption the resistant transcriptome response ET-activated signaling pathway increases during pathogenesis in the resistant host. study shows that the wild accession B. villosa is a novel and unique source of quantitative resistance against the economically important fungal pathogen S. sclerotiorum. We identied genomic regions in B. villosa that will be useful in breeding for improved Sclerotinia-resistance and showed that the genetic basis may only be partially conserved across wild Brassica species. Furthermore, we show that the ET-activated signaling pathway is an essential regulator of the defense response of the resistant B. villosa and associated with an activation of the phytoalexin biosynthetic process and a response to ROS. We point out that the wild Brassica oleracea complex is an important source for resistance breeding in oilseed rape against Sclerotinia. We demonstrated the successful application of the B. napus 15k-SNP-chip-microarray for genetic mapping in the wild B. villosa and propose that a similar strategy in other wild Brassica species with high resistance against Sclerotinia, such as B. drepanensis or B. rupestris, will reveal more QTLs. Furthermore, we think that a similar approach in the Brassica rapa complex may result in the identication of improved resistance in the highly susceptible A-genome (Mei et al. 2011). We conclude that trichomes are non-functional in defense against S. sclerotiorum in the wild B. villosa but are partially co-localized with Sclerotinia-resistance.


Introduction
The necrotrophic fungus Sclerotinia sclerotiorum (Lib.) de Bary, a soil-borne fungal pathogen, causes the sclerotinia stem rot disease (SSR) in its host. The fungus embraces a broad spectrum of host plants with more than 400 species, including many economically important crops (Boland andHall 1994, Bolton et al. 2006). The SSR overwinters as sclerotia in the soil. In spring, apothecia growing on sclerotia carpogenically germinate producing ascospores that are released into air currents and deposited to the petal and stem axil of its host plants. When conditions are favorable, the fungus starts to grow and infect the healthy stem tissue. The pronounced virulence of the fungus is attributed to among others a broad repertoire to produce cell-wall degrading enzymes, phytotoxins, and secreted effector-proteins (Amselem et al. 2011, Derbyshire et al. 2017. Most common symptoms are bleached lesions traversed by white mycelium in the stem or branch tissues and the formation of black sclerotia inside the infected tissue (Bolton et al. 2006). SSR is one of the most destructive diseases in many oilseed rape cultivation areas worldwide. The tremendous increase in cultivation area of oilseed rape in combination with shorter crop rotation cycles favored SSR dispersal over the past decades. Stems of infected plants tend to burst and shatter. The weakened stem stability and resulting lodging of the plants can cause severe yield and quality losses in oilseed rape cultivation (Derbyshire and Denton-Giles 2016). Resistance to Sclerotinia is mainly measured via leaf-, petiole-, or stem-inoculations on the basis of Zhao et al. (2003Zhao et al. ( , 2004. Studies attempting to assess correlations between the different resistance traits reported contradictory results (Mei et al. 2011, Mei et al. 2013, Uloth et al. 2013, You et al. 2016, Taylor et al. 2018 and more effort is needed to determine their genetic link. Though SSR can be effectively controlled by application of fungicides (Derbyshire and Denton-Giles 2016) the increasing restriction of fungicide use due to its potential environmental and health hazards and the emergence of resistant isolates ) ask for alternative control strategies, worldwide.
Breeding for resistant varieties is an important method in plant disease management. But, genetic resistance against SSR is generally lacking in the B. napus gene pool (Derbyshire and Denton-Giles 2016). To date, few genotypes that feature partial SSRresistance are available (Zhao and Meng 2003, Wang et al. 2004, Taylor et al. 2015. Quantitative trait loci (QTL) for SSR-resistance have been reported in various B. napus mapping populations (Zhao and Meng 2003, Zhao et al. 2006, Yin et al. 2010, Wu et al. 2013, Wei et al. 2014, Behla et al. 2017) and genome-wide association studies (GWAS) identi ed single-nucleotide polymorphisms (SNPs) associated with SSR-resistance in numerous B. napus accessions (Gyawali et al. 2016, Wei et al. 2016, Wu et al. 2016. These studies report useful sources of partial SSR-resistance in the narrow oilseed rape gene pool. Efforts have been made to transfer high SSR-resistance from species intercrosses to the primary gene pool of B. napus (Chen et al. 2007, Garg et al. 2010). The B. oleracea complex, including B. incana, B. rupestris, B. insularis, and B. villosa, was identi ed as valuable pool of high SSR-resistance (Mei et al. 2011, Taylor et al. 2018). Mei et al. (2013) identi ed QTLs for SSR-resistance in a mapping population from an interspeci c cross between the wild B. incana (highly resistant) and the cultivated B. oleracea var. alboglabra (susceptible) and partially transferred this resistance via marker assisted selection (MAS) into the B. napus gene pool (Mei et al. 2015, Ding et al. 2019a, Mei et al. 2020. Comparative transcriptome analysis in the wild B. oleracea species linked this resistance to an early perception of the pathogen followed by a rapid release of reactive oxygen species (ROS) modulated by Ca 2+ signaling pathways and an early suppression of Sclerotinia virulence genes (Ding et al. 2019b). These studies highlight the B. oleracea gene pool as important source for introgression of improved resistance to Sclerotinia into the primary gene pool of B. napus. Herein, we present results from phenotypic and genetic analyses on a segregating F 2 -population from an intercross between the wild B. villosa, highly resistant to Sclerotinia, and the wild, susceptible B. oleracea var. oleracea. For the rst time, we report QTLs for Sclerotinia-resistance and trichome-phenotype in the genome of B. villosa. In addition, the comparison with previously identi ed QTLs in the wild B. incana (Mei et al. 2013, Mei et al. 2017) allows for evaluating the resistance mechanisms existing in different Brassica species. Our genetic studies are complemented by an integrated reference-and de novo-based comparative transcriptome pro ling 8 hours post inoculation (hpi), providing unique insights into the early defense response in the resistant B. villosa. We identi ed 58 B. villosaspeci c signi cantly upregulated defense-related genes that are absent in the reference genome of B. oleracea and demonstrate that the distinct activation of the signaling pathways by jasmonic acid (JA) and ethylene (ET) plays a pivotal role in Sclerotiniaresistance and is associated with a strong immune response, a negative regulation of cell death, and an elevated phytoalexin biosynthesis.

Materials And Methods
Plant material and population structure The seeds of B. oleracea (BRA1909) and B. villosa (BRA1896) were obtained from the Institute of Plant Genetics and Crop Plant Research (IPK) in Gatersleben, Germany. A segregating F 2 population with 510 individuals from an interspeci c cross B. villosa x B. oleracea was divided into two mapping populations with 252 and 258 F 2 -individuals, referred to as Population A and B, and cultivated under greenhouse conditions. From Population A, 234 F 2 -plants were once evaluated for Sclerotinia-resistance via the detached leaf-and petiole-assay, from these 187 were selected for genotyping with the B. napus 15k Illumina In nium SNP-chip. From Population B, 258 F 2 -plants were evaluated once via the detached leaf-and petiole-assay. From these, 184 individuals were selected and genotyped. The phenotypic data from the genotyped Population B was re-evaluated with the leaf-and petiole-assay under greenhouse conditions twice. After resistance evaluation, 171 genotyped individuals in Population B were scored according to their trichome-phenotype.

Resistance screening and population comparison
Resistance evaluation was performed with the detached leaf-and petiole-assay (Zhao et al. 2004, Mei et al. 2011. We used a S. sclerotiorum strain originally isolated from an oilseed rape eld in Chongqing, China (Mei et al. 2011). The fungus was cultured on potato-dextrose agar (20 g/l PDB, 15 g/l Bacto agar) plates with a pH of 5.6 at 21°C and transferred to a new PDA plate 2 days before inoculation. At least three plugs with actively growing mycelia were stamped out with a cork borer (d = 0.8 cm) and placed on the detached leaves with the mycelia-site facing the leaf-surface. The 3 rd and 4 th leaves (counted from the apical meristem) and their petioles were used for inoculation. PDA-plugs with actively growing Sclerotinia-mycelium were xated with 1 ml pipette tips on the open cut of the petioles. Detached leaves and petioles were placed in a tray with wetted paper towel placed around open cut surfaces and sealed with foil. Leaf-lesion area and petiole-lesion length were measured two days post inoculation (dpi). The leaflesion area was calculated with the equation I, where 'a' equals the semi-major axis and 'b' indicates the semi-minor axis of a lesion ellipse.  Mei et al. (2017) and classi ed into ve trichome-groups from '0' (completely hairless) to '4' (densely haired). Individuals were scored according to the intensity of the trichome-layer on the leaf and petioles. Each tissue showing visible trichomes was scored with one point (max. two points). A dense trichome-layer as seen on B. villosa was scored with an additional point for each tissue (max. two points).

Trypan blue staining
Detached leaves of B. villosa and B. oleracea were placed in petri dishes, inoculated with PDA-plugs of actively growing Sclerotinia and sealed with Para lm. After 2 dpi, leave tissue of the junction between necrotic and non-necrotic material was hand-dissected in small rectangles, placed into petri dishes and stained with Trypan blue staining solution according to Fernández-Bautista et al. (2016). Samples were visualized and taken with a SteREO Discovery.V20 microscope (Carl Zeiss AG, Oberkochen, Germany), an AxioCam MRc microscope-cam (Carl Zeiss AG, Oberkochen, Germany), and the AxioVision software (v. 4.8.2; Carl Zeiss AG, Oberkochen, Germany).

Statistical analysis
All statistical analyses were performed via the R software (v.3.6.3; R Core Team 2019). Data handling in R was mainly performed with the dplyr package (v.1.0.2; Wickham et al. 2019). Parental lesions in each population were compared via a linear model.
Pearson's correlation analysis was performed between leaf-and petiole-lesions in each inoculation-assay. The association between trichome-phenotype and Sclerotinia-resistance in Population B was analyzed via a one-way ANOVA. Post-hoc multiple comparisons between the lesion mean of each trichome-group to the grand mean of all groups were performed with the multcomp package (Hothorn et al. 2008 Genotyping and genetic map construction Genomic DNA was isolated from leaves following the CTAB protocol (Rogers and Bendich 1985). DNA was resolved in HPLC-H 2 O and DNA quality and concentration was determined on 1 % agarose gel with Lambda-DNA (Thermo Fisher Scienti c, Massachusetts, USA) and the Image Lab Software (Bio-Rad Laboratories, California, USA). Genomic DNA (20 ng/µl) from each plant were loaded onto 96-well plates and send to TraitGenetics (Gatersleben, Germany) for genotyping with the Illumina® In nium BeadChip technology (Illumina, California, USA) and the 15k Brassica chip (TraitGenetics, unpublished). The chip carries a total of 13,714 SNP markers. All SNP-marker were locally searched against the B. oleracea 'TO1000' reference genome (Parkin et al. 2014) retrieved from Ensembl Plants (Bolser et al. 2016) via BLAST+ (v.2.8.1;Altschul et al. 1990, Camacho et al. 2009) with the following options: -evalue 1e-5, -max_target_seqs 2, -max_hsps 1, -outfmt 6. Raw SNP-alleles were transformed into A/B/H-alleles via customwritten python code. The genetic map was constructed with the R/qtl package (v.1.46-2; Broman et al. 2003). Markers with distorted segregation patterns were ltered out based on a p-value < 1e-10 from the implemented chi-square test and arranged into linkage groups (min. LOD ≥ 8; max. recombination fraction = 35 %). The linkage groups were ordered with the implemented Haldane map function (assuming no crossover interference) and the 'geno.crosstab'-function. The linkage groups were assigned to the chromosomes via the best hits from the local BLAST-search.
QTL analysis for Sclerotinia-resistance and trichomes The QTL analysis was performed with the R/qtl package according to the work ow described in Broman and Sen (2009). First, both populations were separately screened for QTLs. The trichome-group in Population B was excluded as covariate in the analysis after no signi cant interaction between Sclerotinia-resistance and the trichome-phenotype was identi ed. The populations were screened for QTLs in three steps. A single-QTL model scan ('scanone'-function) was performed with the Haley-Knott regression (method = 'hk') followed by a scan with a non-parametric model, which considers the rank-based phenotypes (model = 'np'), when the rst scan detected no QTLs. Peak-markers of identi ed QTLs were used as covariates in the single-QTL model to scan for additive and interactive effects of these markers to other loci followed by a two-dimensional QTL-scan considering epistatic effects. A multiple-QTL model was set up according to the identi ed loci from the scans and screened for additional ('addqtl'-function) and interacting ('addint'-function) QTLs. The model was adjusted and nally tted with the forward/backward model selection algorithm with the Haley-Knott method via the 'stepwiseqtl'-function. The effect and amount of explainable phenotypical variance by each QTL was estimated with an ANOVA of the nal multiple-QTL model. QTL-intervals were estimated with the Bayes credible method. QTLs with overlapping intervals were classi ed as common QTL. The signi cance thresholds were determined via genome-scan-adjusted pvalues based on permutation tests (10,000 permutations for the single-QTL scans; 2,000 permutations for the two-dimensional scans). Phenotypes of identi ed QTLs from both populations were transformed into relative lesion values on the basis of Equation II and merged for a ne mapping. Please see the 'code availability statement' for more information.
For differentially expressed gene (DEG) analysis, the gene count matrix was extracted with the enclosed python script in the StringTie software package. Unmapped reads from the reference transcriptome assembly were then extracted from the BAM les via 'samtools view' with the following parameters: -f 12; -F 256. Unmapped BAM les were converted to fastq format via the 'bamtofastq' utility from BEDtools (v.2.29.2; Quinlan and Hall 2010) and re-aligned to the S. sclerotiorum '1980' genome (Amselem et al. 2011). Unmapped reads that neither aligned to B. oleracea nor to S. sclerotiorum were then re-converted to fastq format and assembled de novo via Trinity (v.2.9.0; Grabherr et al. 2011). Counts of the de novo transcripts were estimated via RSEM (v.1.3.3; Li and Dewey 2011). The de novo transcript count matrix was ltered for transcripts with at least 10 counts in each of the three biological replications within a sample. The ltered transcript matrix was added to the gene count matrix from the reference transcriptome assembly. DEG analysis was performed with the merged count matrix via DESeq2 (v.1.26.0; Love et al. 2014). Genes were considered as signi cantly differentially expressed with an adjusted P-value of < 0.05. The calculation of the sample-to-sample distance matrix and the principal component analysis (PCA) are based on the regularized log-transformed counts in DESeq2.
Samples were checked for outliers by Cook's distance. For each gene, the major isoform, calculated by the averaged transcript fragments per kilobase million (FPKM), was used for functional annotation. Graphical illustrations and data conversions were performed in R via ggplot2 (Wickham 2016) and dplyr (Wickham et al. 2019). Please see the 'code availability statement' for more information.

Functional annotation and gene enrichment analysis
The TransDecoder software (v.5.5.0; https://github.com/TransDecoder/TransDecoder/wiki) was used to convert transcript sequences into protein sequences and to identify functional protein-domains. Brie y, longest open reading frame prediction was performed via the 'TransDecoder.LongOrfs' tool. The likely protein-coding regions were used for a homology-based coding region identi cation in Pfam (El-Gebali et al. 2019) via the HMMER software (v.3.2.1; http://hmmer.org/) and in a protein sequence database of Arabidopsis (organism: 3702) downloaded from Uniprot (The Uniprot Consortium 2019) via the BLASTp+ software and the following options: -evalue 1e-5, -max_target_seqs 1, -max_hsps 1, -outfmt 6. The Pfam and BLAST+ results were integrated into the nal coding region prediction via the 'TransDecoder.Predict' tool. Gene ontology (GO) annotations of the Brassica transcripts were retrieved from their closest homolog in Arabidopsis via the database from KOBAS 3.0 (Xie et al. 2011). Additionally, all transcripts were blasted against the B. oleracea (Taxid: 109376) Refseq database. Gene enrichment analysis was performed with the goseq package (v.1.38.0; Young et al. 2010) taking the gene length bias of RNAseq into account. The p-values were adjusted via the false discovery rate method (Benjamini and Hochberg 1995) and GO terms were considered statistically enriched with a FDR ≤ 0.05. The comparative GO analysis was performed and output tables were created with custom-written R-scripts. Heatmaps were created with the ComplexHeatmap package (Gu et al. 2016). Please see the 'code availability statement' for more information.
Quantitative gene expression analysis RNA was isolated at 8 hpi and 16 hpi as described before from three independent biological replications of Sclerotinia-and mock-

Results
Parental performance in the leaf-and petiole-assay B.villosa (BRA1896) was identi ed as highly resistant to Sclerotinia compared to B. oleracea (BRA1909). Differences between the two species were highly signi cant (p < 0.01) in the petiole-assays in both populations and in the leaf-assay in Population A, but not signi cant in the leaf-assay in Population B ( Fig S1). We observed noticeable differences in the fungal spread on infected leaves via Trypan blue staining. A dense and compact structured growth mainly within the necrotic tissue with a sharply delimited junction between healthy and infected tissue was characteristic for the susceptible B. oleracea (Fig. S2). In the resistant B. villosa, the fungal expansion was less structured, mainly centered on the leaf surface with no sharply delimited changeover between healthy and infected tissue and strongly pronounced infection cushions.

Association between leaf-and petiole-lesions
Pearson's correlation analyses revealed a weak signi cant positive correlation between leaf-and petiole-lesions (Population A = p < 0.001; Population B = p < 0.01) explaining about 28 % of variance in Population A and about 2 % of variance in Population B (Fig.  2a). The intersection of F 2 -genotypes classi ed into common categories in the leaf-and petiole-assay was small in both populations ( Fig. S3). Of all individuals, 47 out of 237 'resistant' individuals in either the leaf-or the petiole-assay were commonly classi ed as 'resistant' in both assays. 120 F 2 -plants were commonly classi ed as 'intermediate' and 40 individuals were classi ed as 'susceptible' in both assays. All 207 individuals commonly classi ed as 'resistant', 'intermediate', and 'susceptible' from both populations were selected for genotyping. 164 individuals with mainly moderate variation between the three categories were additionally selected for genotyping. The correlation between leaf-and petiole-lesions in the genotyped individuals increased to 31 % of explainable variance (r = 0.56) in Population A and to 14 % of explainable variance (r = 0.38) in Population B (Fig 2B and C).
Signi cant positive correlations (p < 0.001) were also identi ed in the 2 nd and 3 rd infection of the genotyped individuals in Population B with 25 % (r = 0.50) and 7 % (r = 0.27) of explainable variance, respectively (Fig 2D and E). The correlation increased with respect to the absolute lesion sizes.
Trichome-group '0' was signi cantly associated with higher lesions in the 2 nd leaf-assay (Fig. 4B) and trichome-group '3' was signi cantly associated with increased lesions in the 3 rd petiole-assay (Fig. 4F). A consistent trend between the trichome-groups and their level of Sclerotinia-resistance was not observed.

Construction of a high-dense genetic map
Genotypic data of 371 F 2 -plants was combined to construct a common genetic map for both populations (Table S1). 8,646 SNPs from the Brassica 15k SNP-chip-array were uniquely aligned to the B. oleracea genome via the best hit from the local BLAST-search. The most SNPs (1,466) were aligned to chromosome C03 and the fewest (522) to chromosome C09. Ten F 2 -individuals were discarded for genetic map construction due to considerable genotyping errors. 1,118 SNPs were polymorphic between the parents and ordered into a genetic map with 10 linkage groups with a total length of 784.9 cM and an average distance of 0.7 cM between adjacent markers. Linkage groups were assigned to the chromosomes of B. oleracea according to the local BLAST. Low marker coverage of C04 resulted in two separate linkage groups (C04a, C04b). Genetic positions of markers were concordantly with their assumed physical positions in the B. oleracea genome (Supplementary Data S4).

Identi cation of seven QTLs for Sclerotinia-resistance
Seven QTLs associated with resistance against Sclerotinia were detected with the single-QTL analysis (Table 1). No additional or interacting QTLs were detected with the two-dimensional or the 'stepwiseqtl'-function. The loci are allocated on chromosomes C01, C03, and C07, with two overlapping QTLs (pQTLa, l2QTLb) on C01, two (p1QTLb1, p3QTLb1) on C03, and two (p1QTLb2, p3QTLb2) on C07 representing a common QTL, respectively. pQTLa was detected via the petiole-assay in Population A with the peak at marker Bn-scaff_19564_1-p17934. The marker was mapped to Scaffold01187 (~ 23 kb). Flanking markers were Bn-scaff_15749_1-p118178 (26,828,052 bp) and Bn-scaff_16929_1-p495739 (29,084.454 bp). l2QTLb and l3QTLb were detected in the 2 nd and 3 rd leaf-assay in Population B. The peak of l2QTLb was detected at 51 cM between Bn-scaff_15747_1-p105633 (14,270,425 bp) and Bn-scaff_22790_1-p152675 (16,593,775 bp), which was the marker nearest to the peak. pQTLa and l2QTLb overlapped by ca. ~ 10 cM. Genetic distance between the peaks was 2 cM, corresponding to ca. 12.5 Mbp. The peak of l3QTLb was detected at marker Bn-scaff_16110_1-p976517 (47,351,349 bp) at 79.6 cM. p1QTLb1 and p1QTLb2 were identi ed via the 1 st petiole-assay in Population B and p3QTLb1 and p3QTLb2 were identi ed via the non-parametric method in the 3 rd petiole-assay in Population B. The peak of p1QTLb1 was detected at 12 cM between the markers Bn-scaff_16614_1-p174856 (2,054,448 bp) and Bn-scaff_18936_1-p269153 (3,106,932 bp). Bn-scaff_16069_1-p4306874 (44,016,862 bp) was detected as peak-marker at 54.3 cM for p1QTLb2. p3QTLb1 was detected with a LOD of 3.22 below the LOD-threshold of 3.49 (p = 0.08) with the peak at 13.1 cM at marker Bn-scaff_18936_1-p269153 (3,106,932 bp). Peaks of p1QTLb1 and p3QTLb1 were separated by 1.1 cM and shared the same adjacent peak-marker (Bn-scaff_18936_1-p269153) and therefore declared as common QTL. Bn-scaff_16069_1-p2611780 (42,321,768 bp) was the nearest marker to the peak of p3QTLb2 with a physical distance of 1,695,094 bp to the peak of p1QTLb2 at 44,016,862 bp. l3QTLb showed a partial overlap of 4.6 cM, corresponding to 607.116 bp, at the end of the interval from p1QTLb2 but no overlap with p3QTLb2. The peaks between l3QTLb and p1QTLb2 were separated by 32.63 cM (3,334,487 bp). Two identical non-signi cant peaks were observed in the 1 st and 3 rd petiole-assay on C06 (Fig. S4B). An additional scan with the two-QTL model (y ~ p1QTLb1 + p1QTLb2) in the 1 st petiole-assay increased the LOD from 2.79 to 3.53 below the LOD-threshold of 3.56 (p = 0.053). Both peaks from the 1 st and 3 rd petiole-assay on C06 were localized at 46.8 cM ( Fine mapping of QTLs for Sclerotinia-resistance Similar LOD-pro les between assays in both populations were observed (Fig. S4). Noticeable, several peaks were observed on chromosome C01 adjacent to pQTLa and l2QTLb from leaf-and petiole-assays in Population B and several non-signi cant peaks were observed on C02. The leaf-assay from Population A contributed to one peak on C08, which was not observed in any other assay (Fig. S4A). In order to increase the power and precision to detect QTLs, lesion values were transformed into relative lesions via Equation II with following modi cation: lesions of B. villosa and B. oleracea were replaced by lesion-values of the genotyped F 2plants at the border of the 25 % and 75 % quantile in each assay, respectively. We merged the relative lesion values from the petioleassay (pQTLa) in Population A with the relative lesion values from the 2 nd (l2QTLb) and 3 rd (l3QTLb) leaf-assay, as well as with the relative lesion values from the 1 st (p1QTLb1/p1QTlb2) and 3 rd (p3QTlb1/p3QTLb2) petiole-assay in Population B. The merged scans were performed with the Haley-Knott regression and are summarized in Table 1. The combined analysis increased the LOD and decreased the estimated intervals for all QTLs except for p1QTLb1/p3QTL and l3QTLb. The joint scan of phenotypes from pQTLa and l2QTLb (pAlB2) showed an additive effect that increased the LOD of pQTLa and l2QTLb from 6.2 to 10.7 with a peak at 50 cM (Fig. 5A). The estimated QTL-interval decreased to 7.1 cM (13,002,326 bp -29,084,454 bp). The peak was detected between the anking markers: Bn-scaff_15747_1-p105633 (14,270,425 bp) and Bn-scaff_22790_1-p152675 (16,593,775 bp). Other phenotype combinations did not increase the LOD on chromosome C01. Evidence for QTLs on C03 and C06 vanished completely.
Peaks on chromosome C03 shifted to 91.8 cM (36,288,851 bp) and 88.5 cM (33,853,232 bp) but were not signi cant ( Fig. 5B and C). The joint scan of phenotypes from pQTLa and l3QTLb (pAlB3) decreased the LOD of l3QTLb from 4.07 to 3.77 (Fig. 5D) and moved the peak to 80.5 cM at Bn-scaff_16110_1-p426547 (47,901,219 bp). The estimated QTL-interval increased from 9.16 cM to 14.54 cM (2,367,296 bp). The phenotypes from pQTLa exerted an equal additive effect to the phenotypes from p1QTLb2 (pApB1) and p3QTLb2 (pApB3) (Fig. 5E and F). LOD-scores increased from 4.56 to 5.29 for p1QTLb2 and from 3.78 to 4.41 for p3QTLb2. The peaks moved from 54.3 cM to 58 cM and from 46.9 cM to 59 cM, respectively. Nearest peak-markers were: Bn-scaff_16069_1-p4828119 (44,538,108 bp) and Bn-scaff_16110_1-p3674923 (44,652,843 bp). The estimated interval for p1QTLb2 decreased from 13.14 cM (2,062,672 bp) to 8.8 cM (1,470,303 bp) for pApB1 and increased for p3QTLb2 from 32.41 cM (5,817,550 bp) to 33.75 cM (5,619,756 bp) including the non-signi cant distal peak at the end of C07. The results on chromosome 7 indicate l3QTLb as separate QTL at the end of this linkage group.

Identi cation of QTLs for trichome-phenotypes
Five QTLs associated with trichome-development were identi ed ( Table 1). The single-QTL analysis identi ed three genomic regions on C01 (t1QTLb), C04b (t2QTLb), and on C09 (t3QTLb). A single-scan controlling for the major QTL on chromosome C01 and a two-dimensional QTL-scan indicated two additional loci on C03 (t4QTLb) and C07 (t5QTLb) with additive effects. A scan with the additive multiple-QTL model repeatedly detected the two QTLs with a LOD of 3.58 for t4QTLb and of 5.96 for t5QTLb (Fig. 6). The LOD of t4QTLb was at the level of the LOD-threshold (p = 0.058). Both QTLs were added to the model with additive effects.

Comparative transcriptome pro ling
Hierarchical clustering and PCA analysis of the RNAseq samples showed a grouping according to species and treatment. 58 % and 39 % of variance between the samples were attributed to the species and the response to the treatment, respectively ( Fig. S5A and   B). Overall, 63,995 genes were identi ed in both species. We identi ed 6,630 up-and 1,829 downregulated DEGs in the resistant B. villosa and 7,209 up-and 3,566 downregulated DEGs in the susceptible B. oleracea at 8 hpi with Sclerotinia. Of these, 5,095 up-and 751 downregulated DEGs were common in both species (Fig. 7A). In total, 854 genes had a signi cant difference in their log2-fold change between the species as response to the inoculation. 542 genes had a positive log2-fold change difference, meaning a signi cant stronger induction in the resistant B. villosa, while 312 genes showed a signi cant stronger induction in the susceptible B. oleracea (Fig. 7B). We identi ed 111 GO terms that were commonly enriched in both species that addressed direct pathogen-induced reactions, such as a response to chitin (GO:0010200) and a response to fungus (GO:0009620). In both species, responses to ethylene (ET; GO:0009723) and to abscisic acid (ABA; GO:0009737) were signi cantly stronger enriched than responses to salicylic acid (SA; GO:0009751) and jasmonic acid (JA; GO:0009753). The ET-activated signaling pathway (GO:0009873) was 7 x 10 6 -times more signi cantly enriched in the resistant B. villosa in comparison to B. oleracea. In the susceptible host, we identi ed a 50-times more signi cantly enriched response to SA (GO:0009751) and a three-times more signi cantly enriched response to JA (GO:0009753) in B. oleracea. Both species showed a highly signi cant (log10-FDR < -30) response to decreased oxygen levels (GO:0036293). The immune response (GO:0006955) was strongly addressed in both species, but 7.6 x 10 11 times more signi cantly in the resistant B. villosa. Regulatory processes of the immune system (GO:0050776), a negative regulation of cell death (GO:0060548), and a positive regulation of defense response (GO:0031349), as well as a response to reactive oxygen species (ROS; GO:0000302) were speci cally enriched in B. villosa. The resistant BRA1896 showed overall a strong increase in the primary metabolism (e.g., GO:0015979; GO:0016168; GO:0047899; GO:0009538) and in the phytoalexin metabolic process (GO:0052314; GO:0046217; GO:0052317). The susceptible B. oleracea showed a speci c enrichment of the glycosinolate metabolic process (GO:0019757) and the ABA-activated signaling pathway (GO:0009738). The GO enrichment analysis of the 854 genes with signi cant differences in their log2-fold change expression revealed multifaceted transcriptomic strategies in the commonly enriched terms of both species due to the inoculation. For example, 53 genes associated with the hormone-mediated signaling pathway (GO:0009755) showed a signi cant difference in their expression between the two species as result of the inoculation.
Further differences were identi ed in the signal transduction (GO:0007165; 94 genes), in the response to decreased oxygen levels (GO:0036293; 28 genes), in the response to ABA (GO:0009737; 42 genes), and in the ET-activated signaling pathway (GO:0009873; 17 genes). The complete results of the GO analysis for each species are in the supplementary data (Data S7 -S9). The detailed comparative analysis is available as report (see the 'code availability statement'). The RT-qPCR analysis of the phytohormone markers AOC3 (JA), ETR2 (ET), LOX3 (JA), NCED3 (ABA), PR1 (SA), and PDF1.2 (JA) con rmed the different activation of the hormone-mediated signaling pathways observed in the RNAseq (Fig. S6). AOC3 and LOX3 showed strong signi cant expression at 8 hpi and 16 hpi in the susceptible host, whereas a weaker non-signi cant expression was observed at both time points in B. villosa. Expression of PR1 showed strong variation and was higher in B. villosa, which was also observed in the RNAseq data. ETR2 was highly induced at both time points in the resistant B. villosa and showed an increase from 5.39-fold at 8 hpi to 8.57-fold at 16 hpi. A weaker non-signi cant gene expression was observed in B. oleracea. We observed a signi cant 9.38-fold and 12.1-fold expression of NCED3 at 8 hpi in B. villosa and B. oleracea. The expression decreased to non-signi cant 5.66-fold in B. villosa and increased signi cantly to 18.61-fold at 16 hpi in B. oleracea. PDF1.2 showed a non-signi cant 2.36-fold and 1.14-fold expression at 8 hpi and a decrease to 0.03-fold and 0.44 at 16 hpi in the resistant and the susceptible host, respectively. The qPCR data at 8 hpi generally con rmed our RNAseq data (Fig. S7).

Identi cation of candidate genes in B. villosa
For candidate gene analysis, we focused on the major peak regions from pAlB2 (14,270,425 bp -16,593,775 bp) and pApB1/pApB3 (44,538,108 bp -44,736,714 bp) with left and right anking markers next to the peak. Our transcriptome data revealed 242 genes with 19 and 44 DEGs in the resistant and susceptible host in the major peak area of pAlB2 (Supplementary Data S10). Four of the 242 genes had a signi cant difference in expression. The peak region of pApB1/pApB3 harbors 27 genes of which no gene showed a signi cant difference in expression between B. villosa and B. oleracea (Supplementary Data S11). Six and seven of the 27 genes were identi ed as differentially expressed in B. villosa and B. oleracea, respectively. Gene annotations varied from uncharacterized proteins to kinase like proteins. We did not identify genes with striking expression patterns between both species. Therefore, we set the focus to B. villosa-speci c transcripts from the de novo assembly. Overall, 15,251 de novo transcripts were reconstructed from both species of which we identi ed 3,144 and 3,230 transcripts as DEGs. A comparison of the upregulated transcripts revealed 413 DEGs speci cally upregulated in the resistant B. villosa. Of these, 110 DEGs are associated with response to stress (GO:0006950), 58 DEGs are associated with defense response (GO:0006952), 29 DEGs are related to immune system process (GO:0002376), and 56 DEGs are associated with a response to oxygen-containing compound (GO:1901700). A comparative gene expression analysis of the 58 DEGs associated with defense response revealed that most of these genes are speci c to B. villosa (Fig. 7C, Supplementary Data S12). Nine of these 58 DEGs showed contrary expressions in B. oleracea.

Discussion
The petiole-assay is an e cient and reliable method to assess Sclerotinia-resistance Both assays identi ed B. villosa as highly resistant to Sclerotinia, in accordance with previous observations (Mei et al. 2011, Taylor et al. 2018. The discrepancy observed between the petiole-assays and the leaf-assays is clearly attributed to different inoculation systems. We observed that a fast drying of PDA-plugs on the leaf-surface and irregularities of the leaf-surface severely impeded the inoculation process in some leaf-assay cases. In the petiole-assay, PDA-plugs are well protected from dehydration by the surrounding pipette tip, and the fungus can easily in ltrate the petiole via an open cut of the petiole tissue. A high degree of variation in leaf-architecture and trichome-phenotype was obvious in F 2 -plants. Even within an individual plant, the leaf-lesion variation was higher than that in the petiole-assay. Noticeably, the comparison, of the lesion-distribution between Population A and Population B showed a similar number of 'resistant', 'intermediate', and 'susceptible' F 2 -individuals in the petiole-but not in the leaf-assays. But the expected transgressive segregation was given in the petiole-and leaf-assay of Population A, and in the petiole-assay of Population B, in which the leaf-assay classi ed most plants (120 F 2 -plants) as 'resistant'. This re ects an overestimated proportion of 'resistant' F 2 -individuals by the leaf-assay in Population B. Though the leaf inoculation is a well-established Sclerotinia-inoculation technique (Zhao and Meng 2003, Mei et al. 2011, Wu et al. 2013, Joshi et al. 2016, our data suggest that the petiole-assay (Zhao et al. 2004) is a more reliable, reproducible, and e cient tool for a large-scale screening for Sclerotinia-resistance. In consistence, Taylor et al. (2018) found no signi cant correlation between leaf-and stem-resistance in a set of wild Brassica species. Uloth et al. (2013) and You et al. (2016) found no association between leaf-and stem-resistance under eld-conditions in diverse Brassica species. By contrast, a positive correlation between leaf-and stem-resistance under eld-and controlled-environments by arti cial inoculation was reported by Mei et al. (2011Mei et al. ( , 2013. Even though several factors can impair the leaf-and petiole-inoculation systems, an additive effect of the merged leaf-and petiole-phenotypes from Population A and B for the major QTL on C01 observed in this study strongly supports for a common genetic basis determining the leaf-and petiole-resistance in the wild B. villosa. A low power of the leaf-assay might be the cause for no other overlapping QTLs between both assays.

B.villosa is a unique source of Sclerotinia-resistance
The separate analysis of two mapping populations facilitated the identi cation of a minor QTL on chromosome C03 in Population B. In total, seven loci associated with Sclerotinia-resistance were identi ed in the genome of the wild B. villosa. The merged analysis of phenotypes from Population A and B, enabled a ne mapping of the QTLs on chromosomes C01 and C07. The phenotypes for pQTLa and l2QTLb showed a strong additive effect when combined with an increase of the LOD from 6.2 to 10.7. The QTL-interval was narrowed down from 28 Mbp (l2QTlb) and 26 Mbp (pQTLa) to 16 Mbp and the peak was localized between 14,270,425 bp and 16,593,775 bp. For the QTL on chromosome C07, the ne mapping consequently reduced the distance between the peaks from p1QTLb2 and p3QTLb2 from 1,695,094 bp to 114,735 bp. An integrated analysis conducted by Li et al. (2015) physically mapped QTLs for Sclerotinia-resistance from several studies to the genome of B. napus. Most conserved QTLs for Sclerotinia-resistance in B. napus are located on chromosome C06 and C09. In this study, we got low evidence for a minor QTL only on C06 in Population B but could not validate it. This may because of different inoculation systems, materials, and markers used in the studies. In similar studies, Mei et al. (2013Mei et al. ( , 2017 produced and analyzed a mapping population from a cross between wild B. incana (resistant) and the cultivated B. oleracea var. alboglabra (susceptible). A major QTL for Sclerotinia-resistance on chromosome C09, which explain about 13.6 % of phenotypical variance, and a QTL on chromosome C01 were reported. Remarkably, the QTL on chromosome C01 in B. oleracea, ranging from 13 Mbp to 29 Mbp identi ed in this study perfectly matches the previously reported QTL from Mei et al. (2013) on chromosome C01 in B. napus . Thus, comparison between the two QTL as well as their physical overlapping regions may provide more insights into the genetic basis for Sclerotinia-resistance. The identi ed major QTL on C09 and the minor QTL on C07 detected in B. incana might support the hypothesis of different resistance mechanisms existing in different wild Brassica species. Our recent data support that the wild Brassica species B. drepanensis, B. rupestris, and B. insularis (data not shown) are valuable genetic sources for breeding for resistance against SSR, which can be partially transferred into the B. napus gene pool with MAS as demonstrated by Mei et al. (2015).
Trichome-phenotype is partially co-localized but not associated with Sclerotinia-resistance Non-glandular trichomes act mainly as physical barriers towards biotic stresses, whereas glandular trichomes can secrete chemicals with antifungal activity (Hauser 2014). It was shown that non-glandular trichomes of Brassica villosa subsp. drepanensis accumulated high levels of metals (Nayidu et al. 2014a) but their role in plant resistance against Sclerotinia is unknown. The fact that the petiole-assay clearly distinguish resistant and susceptible parental plants implies that trichomes do not play an essential role in plant resistance. Microscopic images suggest that trichomes in B. villosa seem to be anchor points for the mycelia on the surface rather than being a barrier to the fungus. In support of this, we did not nd a consistent trend between decreasing lesionsizes and increasing trichome-phenotype with one exception, in which smaller leaf-and petiole-lesions were given in the densely haired trichome-group '4' once. In one petiole-assay, hairy F 2 -individuals of trichome-group '3' were to contrast signi cantly more susceptible to the Sclerotinia infection as compared to the grand mean of all groups. Interestingly, we found that the genetic region of t1QTLb for trichomes was overlapping with that of the major QTL for Sclerotinia-resistance on chromosome C01. Peaks of both QTLs were detected between the markers Bn-scaff_15747_1-p105633 (14,270,425 bp) and Bn-scaff_15747_1-p105633 (16,593,775 bp). It is to mention that the smaller lesion values in the trichome-group '4' co-occurred with l2QTLb2 in Population B, which showed also the strongest correlation between leaf-and petiole-lesions. Thus, we believe that these loci are partially co-segregating but the link is masked by low effect-sizes, such as in the petiole-assay. Following this, we conclude that trichomes are not an essential element contributing to Sclerotinia-resistances in the wild B. villosa, though they are partially co-localized on chromosome C01.

Trichome-phenotypes are partially conserved in wild Brassica species
In A. thaliana, it has been shown that trichome-morphology is regulated by a complex network of multiple transcription factors, including the positive enhancers GL1, GL2, GL3, EGL3, TTG1, and the negative regulator TRY (Balkunde et al. 2010). Expression analysis of these orthologous genes in the closely related B. villosa subsp. drepanensis identi ed three copies of the TRY gene allocated to chromosomes C02, C03, and C09 (Nayidu et al. 2014b). We identi ed ve orthologous genes of TRY from A. thaliana (At5g53200) in our RNAseq data, of which one copy (Bo1g051040) is located at 14,433,200 bp on C01 in the B. oleracea 'TO1000' genome within the t1QTLb peak-marker Bn-scaff_22790_1-p152675 (16,593,775 bp) and the anking marker Bn-scaff_15747_1-p105633 (14,270,425 bp). The gene was weakly present in B. oleracea (FPKM = 0.129) in the inoculated sample. One copy of GL1 (Bo7g090950) and TTG1 (Bo7g096780) are located on chromosome C07 within the physical interval of t5QTLb (27,506,654 bp -40,491,426 bp). Only TTG1 was present in our RNAseq data and showed no noticeable difference in abundance between both species. We identi ed two copies of EGL3 (Bo9g029320; Unigene.32857) both located on chromosome C09 out of the physical interval of t3QTLb. No difference in gene expression was observed between the glabrous and the hairy species. Our data suggests that these copies may not be involved in the trichome-phenotype in our B. villosa species. Nayidu et al. (2014b) measured a high gene expression of the negative regulator TRY in their wild B. villosa species and identi ed one copy (TRY-1) with higher transcript abundance compared to the other two copies by RNAseq. This is contradictory to the known model in A. thaliana in which TRY inhibits trichome-development (Hülskamp et al. 1994). Further, we did not identi y any gene expression of the TRY locus in our B. villosa species and a very low gene abundance in the glabrous B. oleracea. Mei et al. (2017) reported a 3:1 segregation pattern between glabrous and hairy plants in 1063 F 2 -plants from their mapping population and identi ed one major QTL for leaf-trichomes on chromosome C01 with the TRY (Bol013124) gene as major candidate within the QTL. The gene was successfully used as functional cleaved ampli ed polymorphic site (CAPS) marker for the selection of glabrous and hairy individuals in F 2:3 families. In contrast, in our study a signi cant deviation from a 3:1 segregation was given and the alleles from the glabrous B. oleracea were dominant in the major QTL on C01. A direct comparison to our study is di cult since Mei et al. (2017) used the B. oleracea var. capitata reference genome, but a DNA sequence alignment of these two genes revealed noticeable structural differences and it is unclear whether Bol013124 and Bo1g051040 are two distinct genes or allelic variants and if other orthologous genes of TRY, GL1, TTG1, and EGL3 may play an important role in our B. villosa species. It is reasonable to assume that the QTL identi ed by Mei et al. (2017) and our QTL are a conserved locus in the wild Brassica species. The discrepancy between the segregation patterns in these two studies might be caused technically rather than genetically. Further analyses, including the gene expression analysis of the TRY locus and the functional validation of the CAPS marker in our mapping populations may shed more light on molecular mechanisms of trichome-phenotypes observed in B. villosa and B. incana.
Differential activation of the ET-/JA-activated signaling pathways Our data shows that ET and ABA were the most addressed phytohormones in response to the Sclerotinia-inoculation. Though a signi cant response of the ET-activated signaling pathway was detected in both species (BRA1896 = -12.05 log 10 -FDR; BRA1909 = -5.2 log 10 -FDR), it was remarkably stronger enriched in the resistant B. villosa. We further identi ed 17 signi cantly differentially regulated downstream ET response transcription factors (ERF) between B. villosa and B. oleracea as a result of the Sclerotiniainoculation. 14 of these ERFs were positively induced in B. villosa. These different expression pro les and the increase of expression of the ethylene receptor gene ETR2 in B. villosa at 16 hpi support the assumption that the resistant transcriptome response is distinctly regulated through the ET-activated signaling pathway and even increases during pathogenesis in the resistant host. Interestingly, we observed higher gene expression pro les of the JA-biosynthesis markers AOC3 and LOX3 in the susceptible B. oleracea, which is in accordance with the RNAseq that showed a stronger signi cant response to JA in this species. Thus, we also assume a distinct activation of the JA-mediated signaling pathway in the two species. The ET-and JA-mediated signaling pathways are key components in regulating plant defense to necrotrophic pathogens and synergize the ERF branch via ERF1/ORA59 (Pré et al. 2008, Broekgaarden et al. 2015. A key marker gene of the ERF branch is PDF1.2 which is regulated by ORA59, an essential integrator of the ET-and JA-signal transduction pathway (Pré et al. 2008). Our data showed signi cant expression of one ORA59-like gene (Bo8g114710) in B. villosa but no signi cant induction of PDF1.2. Therefore, we assume that the ET-and JA-signaling pathways are not synergize the common ERF branch and activate different resistance mechanisms in both species. Yang et al. (2017) demonstrated that an infection of resistant rice cultivars with the blast fungus M. oryzae activated the ET-signaling pathway resulting in an increase of ROS and phytoalexin production which is in accordance with our observed signi cant enrichment of the phytoalexin biosynthetic process and a speci c response to ROS in the resistant B. villosa. Conversely, the susceptible B. oleracea showed a strong response of the glucosinolate and sulfur compound metabolic process also observed in B. napus in response to Sclerotinia and might be a result of the JA-mediated pathway (Wei et al. 2016). Strikingly, the ABA-activated signaling pathway was speci cally addressed in the susceptible B. oleracea (BRA1896 = -0.13 log 10 -FDR; BRA1909 = -1.95 log 10 -FDR) in which 19 ABAassociated genes showed a signi cant difference in expression between both species in response to the Sclerotinia-inoculation. The JA-mediated MYC branch, antagonistic to the ERF branch, is responsible for defense against herbivores and co-regulated by ABA , Broekgaarden et al. 2015. However, the orthologue Bo2g159220 of the MYC-marker gene VSP2 was downregulated in B. oleracea. An antagonistic suppression of the JA-activated pathway by SA might be indicated by a higher expression of PR1 in B. villosa compared to B. oleracea. But a more detailed investigation on the crosstalk of the hormone-mediated signalling pathways is needed as expressions of PR1 and PDF1.2 were vague in both species.
Integrated transcriptome analysis reveals novel B. villosa-speci c genes The identi cation of promising candidates from the QTL is limited to genes that are present in the 'TO1000' reference genome. An analysis of the reference genes in the peak regions of chromosome C01 and C07 revealed 242 and 27 genes, respectively. Gene expression pro les showed subtle differences further impeding the selection of candidates. We hypothesized that important B. villosa-speci c genes are absent in the 'TO1000' reference genome and performed an integrated reference-and de novo-based RNAseq analysis. Hence, we were able to identify 413 upregulated B. villosa-speci c genes in response to the Sclerotinia-inoculation. By a comparative functional analysis, 58 candidate genes were identi ed that are associated with a defense response but are lacking a genetic link to the here identi ed QTLs. Genomic re-sequencing of B. villosa may facilitate a linkage of these genes to the QTLs and allow the selection of promising candidates for further studies.

Conclusion
Our study shows that the wild accession B. villosa is a novel and unique source of quantitative resistance against the economically important fungal pathogen S. sclerotiorum. We identi ed genomic regions in B. villosa that will be useful in breeding for improved Sclerotinia-resistance and showed that the genetic basis may only be partially conserved across wild Brassica species. Furthermore, we show that the ET-activated signaling pathway is an essential regulator of the defense response of the resistant B. villosa and associated with an activation of the phytoalexin biosynthetic process and a response to ROS. We point out that the wild Brassica oleracea complex is an important source for resistance breeding in oilseed rape against Sclerotinia. We demonstrated the successful application of the B. napus 15k-SNP-chip-microarray for genetic mapping in the wild B. villosa and propose that a similar strategy in other wild Brassica species with high resistance against Sclerotinia, such as B. drepanensis or B. rupestris, will reveal more QTLs. Furthermore, we think that a similar approach in the Brassica rapa complex may result in the identi cation of improved resistance in the highly susceptible A-genome (Mei et al. 2011). We conclude that trichomes are non-functional in defense against S. sclerotiorum in the wild B. villosa but are partially co-localized with Sclerotinia-resistance.

Declarations
Code availability The main code and results are available as reports at http://doi.org/10.5281/zenodo.4556803.

Contributions
TB conducted main experiments and constructed the genetic map, designed and conducted the RNAseq, qPCR experiments, and bioinformatics analyses, and drafted the manuscript. JM contributed to generation of population, phenotyping and genotyping experiments. MS supported phenotyping and computational analysis. WY supported generation of population, phenotyping and genotyping experiments. MH supported the statistical analysis. SR and GL provided plant materials and supported the project. DG conceived and directed the project and nalized the manuscript. All authors read and approved the nal manuscript. QTL were labeled by trait ('p' = petiole, 'l' = leaf, and 't' = trichomes) with numbers representing the replication, by mapping population ('a' = Population A, 'b' = Population B) followed by a second number to distinguish multiple QTL from one assay. Merged QTL were labeled by trait and mapping population (in capitals) and numbers representing the replication of the assay. a Marker at peak or nearest to the peak.  Trypan blue. In the susceptible accession BRA1909, the fungal tends to a dense and compact structured growth mainly within the necrotic tissue with a delimited junction between healthy and infected tissue indicated by a layer of dead plant cells in front of the hyphae. In the resistant BRA1896, fungal expansion appears less structured, focusing mainly on the leaf surface with no delimited junction between healthy and infected tissue.