Population-level whole-genome sequencing of Ascochyta rabiei identifies genomic loci associated with isolate aggressiveness Authors: Niloofar Vaghefi, Ido Bar, Jonathan Wanderley Lawley, Prabhakaran Thanjavur Sambasivam, Melody Christie, Rebecca Ford Published: 2024 ABSTRACT AbstractAscochyta blight caused by the ascomycete Ascochyta rabiei poses a major biotic threat to chickpea (Cicer arietinum) industries worldwide and incurs substantial costs to the Australian multimillion-dollar chickpea industry in both disease control and yield loss. The fungus was introduced to Australia in the 1970s from an unknown source population and, within a few decades, successfully established in all Australian agroecological chickpea-growing regions. Although genetically highly clonal, a broad range of phenotypic variation in terms of aggressiveness exists among the Australian A. rabiei isolates. More recently, highly aggressive isolates capable of causing severe disease symptoms on moderate to highly resistant chickpea cultivars have increased in frequency. To identify genetic loci potentially associated with A. rabiei aggressiveness on Australian chickpea cultivars, we performed deep genome sequencing of 230 isolates collected from a range of agroecological chickpea-growing regions between 2013 and 2020. Population genetic analyses using genome-wide SNP data identified three main clusters of genetically closely related isolates in Australia. Phylogenetic analyses showed that highly aggressive phenotypes developed multiple times independently throughout the phylogeny. The results point to a minor contribution of multiple genetic regions and most likely epigenomic variations to aggressiveness of A. rabiei isolates on Australian chickpea cultivars. Impact Statement This research introduces new knowledge on the Australian Ascochyta rabiei population structure, molecular pathogenicity drivers and evolution as a clonal pathogen through a comprehensive whole-genome sequencing approach. The knowledge generated on the structure and origin of the Australian A. rabiei and the existence of only one mating type continue to inform researchers, growers, breeders and the broader industry on the importance of continued tight biosecurity measures and inform the development of accurate and informed disease management and resistance breeding strategies. This research provides a rare real-life example to the effect of genetic drift on a clonal pathogen population and the importance of biosecurity to protect introduction from the non-endemic isolates through seed importation in the current era of international markets. Data Summary An online dataset containing the data and code required to reproduce the results found in this publication was deposited in Zenodo (DOI: 10.5281/zenodo.12575659). Isolate aggressiveness and collection metadata are available in the Ascochyta dashboard at http://bit.ly/asco-dashboard. Raw sequencing data used in this study were deposited in the National Center for Biotechnology Information Short Read Archive and were available through BioProject PRJNA1175002. The authors confirm that all supporting data, code and protocols have been provided within the article or through supplementary data files. Introduction Diseases caused by invasive agricultural pathogens, and specifically necrotrophic fungi, present a massive threat to food production industries. Of these, Ascochyta blight caused by Ascochyta rabiei poses a major biotic threat to chickpea (Cicer arietinum) worldwide and costs the Australian multimillion-dollar chickpea industry an estimated $34.9 million in disease control expenses and $4.8 million in yield losses [1]. The fungus was introduced to Australia in the 1970s from an unknown source population and, within a few decades, successfully established in all Australian agroecological chickpea-growing regions where management is heavily reliant on a combination of fungicide application and host resistance [24]. Meanwhile, increases in aggressiveness within A. rabiei populations have resulted in the breakdown of resistance in many chickpea varieties, limiting the efficacy of their host resistance [35]. Within Australia, an extensive annual A. rabiei sampling from all major chickpea-growing regions and subsequent phenotypic assessment have revealed temporal increases in population aggressiveness [6], measured as increases in quantitative pathogenicity, i.e. the severity of symptoms caused by the pathogen, or pathogenic fitness [78]. This is driving the ongoing search for resistance sources within wider germplasm collections [910] and the exploration into the genetic factors and genomic regions underpinning the resistance mechanisms [11]. Although highly clonal, a broad range in aggressiveness exists among the Australian isolates [34 12 13], and highly aggressive isolates, able to cause severe disease on recently released ‘resistant’ chickpea varieties, are increasing in frequency [614]. However, the biological and genetic factors underlying this apparent rapid evolution of isolate aggressiveness are unknown. The clonal nature of the Australian A. rabiei population contrasts with that in most other regions of the world, where A. rabiei populations are genetically diverse and regularly undergo sexual recombination in a heterothallic manner. This is due to the presence of isolates with alternate mating-type idiomorphs (MAT1-1 and MAT1-2) within the population [1518]. However, after the introduction to Australia, A. rabiei underwent a severe bottleneck and reduction in genetic diversity [1319]. Moreover, it was forced into asexuality, likely due to the introduction of a single mating type (MAT1-2 [20]). An intensive infield sampling of diseased crops across chickpea-growing regions and screening of A. rabiei populations over the past decade has identified only the MAT1-2 locus in the Australian population [613 14]. Therefore, all evidence to date points to a genetically homogenous and stable clonal population with low genetic diversity and no signs of recombination in Australia. On the other hand, the increased aggressiveness indicates that the Australian A. rabiei population is evolving rapidly, with more frequent emergence of highly aggressive isolates, despite its asexuality and very limited genetic diversity [4]. Asexual organisms with low genetic diversity are expected to have a slow rate of evolution due to reduced genetic variation, accumulation of deleterious mutations and clonal interference [21]. However, many asexual filamentous fungi are able to generate de novo genetic variation via various other mechanisms, including parasexuality (a nonsexual mechanism of transferring genetic material without meiosis) [22], chromosomal rearrangements [23] and conditionally dispensable chromosomes [24]. However, these mechanisms have not been reported in A. rabiei. Rather, the homogenous population structure, high levels of clonality and the lack of temporal changes in the Australian A. rabiei population are congruent with the lack of sexual reproduction and slow generation of genetic variation. The recent emergence of isolates with high aggressiveness on previously resistant chickpea cultivars is, therefore, unexpected and potentially due to strong selective forces on the standing genetic variation, rather than generation of de novo variation in aggressiveness. In any case, significant differences in the aggressiveness of isolates that show little genetic variation are interesting and warrant further investigation. The most recent Australian A. rabiei population genomics study used DArTseq™ genotyping to generate a set of SNP markers from a reduced representation of the A. rabiei genome for the analysis of genetic relatedness among isolates, to assess population structuring and to identify genetic regions associated with aggressiveness [6]. The discovery of only 212 polymorphic SNP loci among 180 isolates collected from diverse growing regions in Australia further highlighted the clonality of the Australian A. rabiei population and demonstrated the need for a higher marker density to obtain accurate marker–phenotype associations. In the current study, we hypothesized that the increased frequency of highly aggressive isolates in the Australian A. rabiei population is the result of minor genetic modifications, which alter the structure or expression of the existing effector loci or create novel ones to affect aggressiveness traits. To test this, we conducted deep genome sequencing of 230 A. rabiei isolates collected from a range of agroecological zones between 2013 and 2020 to (i) obtain a high-resolution genetic marker set to better characterize the Australian A. rabiei population structure and (ii) identify genetic regions potentially associated with A. rabiei aggressiveness in Australia. Methods A. rabiei isolates, species identification and mating-type screening A total of 230 Australian A. rabiei isolates collected from commercial chickpea fields or naturally infected experimental trial sites from 2013 to 2020 were included in this study (Table S1, available in the online Supplementary Material). Most of the isolates (n=193) were collected in 2020 from chickpea-growing regions in ten agroecological zones with distinct climate and cropping systems (Fig. 1a, Table 1). These were collected using two sampling approaches. First, in fields with high disease incidence, structured, intensive sample collection was conducted in a hierarchical manner, collecting symptomatic chickpea tissue from the four corners and one central location within each field. Second, opportunistic sampling was conducted in the fields where disease incidence was low by haphazardly collecting symptomatic chickpea leaves where the disease was found. Fungal isolations from the 2020 samples were conducted according to Christie and Moore [25], and single-conidium cultures were established on potato dextrose agar (Amyl Media). Additional isolates (n=37) were collected from 2013 to 2017 as reported in earlier studies [613 14]. Metadata on all the isolates included in this study are provided in Table S1. (a) Sampling locations of 230 A. rabiei isolates sequenced in this study. Geographical regions with different colours represent the Australian agroecological zones with distinct climate and cropping systems, as defined by the Australian Grains Research and Development Corporation. Names for the ten agroecological zones from which A. rabiei isolates were collected are provided. Circles depict the frequency of isolates in each agroecological zone from each of the three clusters identified via DAPC, with green, orange and purple representing clusters 1, 2 and 3, respectively. (b) DAPC analysis for A. rabiei isolates based on 2554 neutral genome-wide SNPs. The graph on the left shows BIC plotted against the number of inferred clusters. The ordination plot on the right shows the three assigned clusters of genetically related A. rabiei isolates. The graph at the bottom shows the membership probability of A. rabiei isolates to the three genetic clusters detected through DAPC analysis. Each bar on the horizontal axis indicates an individual, the vertical axis shows the probability of membership and the colours represent the membership of individuals in the predicted clusters. Australian state and territories abbreviations: New South Wales (NSW); Queensland (Qld); Victoria (Vic); South Australia (SA); Western Australia (WA). Table 1.Number of Australian A. rabiei isolates used for whole-genome genotyping by year of collection and agroecological zone*YearNSW1NSW/Qld1NSW/Qld2NSW/VicQldSA1SA/Vic1SA/Vic2WA1WA2 2013 –––––3–––– 2014 –3–––––1–– 2015 –––––4–5–– 2016 ––2––224–– 2017 13–––121–– 2020 10332037421933520*Agroecological zones include NSW1,: NSW Central;, NSW/Qld1,: NSW NE/Qld SE;, NSW/Qld2,: NSW NW/Qld SW;, NSW/Vic,: NSW Victoria slopes;, Qld,: Queensland Central;, SA1,: SA Midnorth-Lower Yorke Eyre;, SA/Vic1,: SA Vic Bordertown-Wimmera;, SA/Vic2,: SA Vic Mallee;, WA1,: WA Central;, WA2,: WA Sandplain (Fig. 1). Australian state and territories abbreviations: New South Wales (NSW); Queensland (Qld); Victoria (Vic); South Australia (SA); Western Australia (WA).NSWNew South WalesQldQuenslandSASouth AustraliaVicVictoriaWAWestern Australia Number of Australian A. rabiei isolates used for whole-genome genotyping by year of collection and agroecological zone* Agroecological zones include NSW1,: NSW Central;, NSW/Qld1,: NSW NE/Qld SE;, NSW/Qld2,: NSW NW/Qld SW;, NSW/Vic,: NSW Victoria slopes;, Qld,: Queensland Central;, SA1,: SA Midnorth-Lower Yorke Eyre;, SA/Vic1,: SA Vic Bordertown-Wimmera;, SA/Vic2,: SA Vic Mallee;, WA1,: WA Central;, WA2,: WA Sandplain (Fig. 1). Australian state and territories abbreviations: New South Wales (NSW); Queensland (Qld); Victoria (Vic); South Australia (SA); Western Australia (WA). NSWNew South WalesQldQuenslandSASouth AustraliaVicVictoriaWAWestern Australia For DNA extraction, single-conidium isolates were grown in 20 ml Czapek-Dox broth (Oxoid, Australia) in 50-ml Falcon tubes (Techno Plas, Australia) on a shaker (Ratek orbital shaker incubator) at 180 r.p.m. at room temperature. Two-week-old mycelia were harvested, dried on autoclaved paper towel, flash-frozen in liquid nitrogen and lyophilized for 48 h. The lyophilized mycelium of each isolate was ground to a fine powder in a 2-ml microcentrifuge tube using stainless steel beads (2.8 mm diameter, Sigma-Aldrich) in a FastPrep-24 kit (MP Biomedicals) (6.5 m s−1 for 15 s). Genomic DNA was extracted from each isolate using a DNeasy Plant Mini kit (Qiagen) according to the manufacturer’s instructions, with the following modifications. Buffer AP1 was pre-heated to 65 °C before 450 µl buffer AP1 was added to the tubes on ground frozen tissue, which were subsequently incubated at 65 °C for 15 min. Also, 160 µl buffer P3 was added to the tubes and incubated on ice for 10 min. Finally, DNA was eluted twice in 35 µl filter-sterilized 10 mM Tris-HCl (pH 8.5). The DNA was quantified using a Qubit dsDNA Broad Range kit (Qubit 3.0, Invitrogen), and the integrity was assessed by agarose gel electrophoresis (1% wt/vol agarose in Tris-acetate-EDTA) amended with 1:20 000 (v/v) nucleic acid stain GelRed (Biotium). The isolates were confirmed as A. rabiei and their mating type was assigned using the A. rabiei-specific multiplex PCR assay described by Barve et al. [15]. Plant materials, inoculum and bioassay conditions A differential host set of chickpea genotypes with industry-established disease reactions in 2021, that is, ICC3996 (moderately resistant), Genesis 090 (moderately susceptible), PBA Seamer (moderately resistant) and PBA HatTrick (moderately susceptible), was used to assess isolate aggressiveness (note that these ratings have since changed; see the current disease ratings at the Grains Research and Development Corporation National Variety Trials Disease Ratings). The highly susceptible cultivar Kyabra was used as a susceptible check in the bioassays. The seeds were provided by the National Chickpea Breeding Program, Tamworth, New South Wales (NSW), Australia. For disease severity bioassays, seedlings were grown in 15-cm diameter pots containing commercial-grade potting mix (Richgro premium mix), fertilized with Nitrosol®, Amsgrow® (4.5 ml l−1) every 2 weeks and watered as required. Plants were grown and maintained at 23±2 °C under an 8-h night/16-h day photoperiod with near-UV light irradiation of 350–400 nm (supplied by Valoya LED Grow Lights, Finland) at Griffith University, Nathan Campus, Queensland, Australia. Single-spored Australian A. rabiei isolates were cultured on V8 juice agar plates and maintained at 22±2 °C with 12-h night/12-h day near-UV light irradiation (350–400 nm) for 14 days. Subsequently, spore suspensions were prepared by harvesting pycnidiospores with sterile water, filtering through sterile cheese cloth (250 mm), and the final concentration of inoculum was adjusted to 1×105 spores per millilitre using a haemocytometer. Two to three drops of Tween 20 (0.02% v/v) were added per 100 ml of spore suspension as a surfactant. Bioassays were conducted in a completely randomized design with two replicates for each genotype×isolate combination and five seedlings per pot in each replicate. Inoculation was carried out using the modified mini-dome technique [2627]. Seedlings were misted to run-off with a single spore inoculum and placed randomly in dark and humid sealed 20-l plastic crates (one crate/isolate). At 24-h post-inoculation (hpi), crate lids were removed, and the seedlings were placed back into the 23±2 °C and 8 h night/16-h day growing conditions until tissue sample collection and disease assessment [28]. Humidity was maintained at 70–75% by misting every 2 days or as needed. Assessment of host reaction and assignment of isolate aggressiveness The disease reaction of each isolate on each of the host genotypes was assessed using the qualitative 1–9 scale proposed by Singh [29] at 21 days post-inoculation where scores of 1 or 3 represented ‘low disease severity’, 5 represented ‘moderate disease severity’ without significant stem infection and 7 or 9 represented ‘high disease severity’ with stem lesions that led to an impact on transpiration, photosynthesis and possible stem breakage. The isolates able to produce a leaf score of at least 7 on >80% and a stem score of at least 7 on >10% of the seedlings of any of the host genotypes assessed were classified as ‘highly aggressive’. Classification into pathogenicity groups (PGs) was determined by the following criteria: low disease response on all hosts (except Kyabra, which is the susceptible check), PG0; high on PBA HatTrick and low on ICC3996 and Genesis090, PG1; high on PBA HatTrick and a combination of low and moderate on ICC3996 and Genesis090, PG2; high on PBA HatTrick and moderate on both ICC3996 and Genesis090, PG3; high on PBA HatTrick and Genesis090 and moderate on ICC3996, PG4; and high on ICC3996, Genesis090 and PBA HatTrick, PG5, as described by Bar et al. [6]. Isolate aggressiveness and collection metadata were deposited in the Ascochyta dashboard and are available at https://bit.ly/asco-dashboard. Whole-genome sequencing and variant calling Whole-genome sequencing was conducted at the Australian Genome Research Facility (Melbourne) using Illumina Nextera library preparation and NextSeq sequencing platform. To assess the reproducibility of variant calling and allow for stringent filtering of the raw data, three DNA samples (AR0039, AR0052 and AR0242) were replicated within and across plates as technical controls. Whole-genome sequencing data processing, mapping and reference-based variant discovery were performed on the Griffith University High-Performance Computing Cluster ‘Gowonda’. Raw sequencing reads were pre-processed to remove adaptor sequences and to trim low-quality bases from the raw reads using fastp v0.20.1 [30] with default parameters. Trimmed reads were aligned to A. rabiei reference genome, strain ArME14 [National Center for Biotechnology Information (NCBI) accession GCA_004011695.2 [31]] using bwa-mem v0.7.17-r1188 [32], followed by several commands of Samtools v1.12 [33], Sambamba v1.0.0 [34] and Picard v2.27.5 [35] to specify a Read Group for each sample, mark duplicates and convert the alignments into sorted coordinates indexed BAM files, respectively. The alignment files were then processed using Freebayes v1.3.5 [36] to call single-nucleotide variants (SNVs) and insertion/deletions (INDELs) from all samples. Quality assessment and filtering Quality filtering of the variant call format (VCF) file produced by Freebayes was conducted using vcfTools v0.1.16 [37] and SnpSift v4.3 [38] on the Linux cluster at the Cornell University BioHPC computing cloud (Ithaca, New York) in multiple steps. First, the dataset was filtered using vcfTools to retain only high-quality genotype calls (GTs) (defined as the assignment of alleles at each site for each individual) with a minimum depth of five and minimum quality of 20 (--minDP 5 --minGQ 20 --recode --recode-INFO-all). The INFO column was recoded every time the data were filtered in vcfTools (--recode --recode-INFO-all) to retain all the INFO in the original files and re-estimate INFO values after applying the specified filters. Sites with more than twice the total expected read depth were removed using SnpSift to exclude potential errors due to mapping/assembly errors in repetitive regions (filter ‘(DP <42000)’). Only biallelic SNVs were retained, and sites with more than 50% missing data and average quality <30 were removed (--min-alleles 2 --max-alleles 2 --max-missing 0.5 --minQ 30 --recode --recode-INFO-all). Individual levels of missing data were estimated using vcfTools (--missing-indv), which was estimated to be <10% for all individuals; thus, no isolates were filtered out. Sites were filtered for polymorphism, average depth and missingness, and only those with a minor allele frequency of 0.01 or greater, average depth of 20 and missing data<10% were retained (--min-meanDP 20 --max-missing 0.9 --maf 0.01 --recode --recode-INFO-all). The entire dataset was subsequently filtered using SnpSift, removing any sites with different GT between the replicated DNA samples to ensure that only high-quality loci with reproducible results were retained. The three replicated technical controls were subsequently removed from the dataset after establishing that bitwise genetic distance estimated for replicated DNA samples was zero. Only polymorphic sites were retained using a polymorphism threshold of 1% (minor allele frequency=0.01) (--remove list.txt --maf 0.01 --recode --recode-INFO-all). The polymorphic SNVs are hereafter referred to as SNPs. In order to investigate the location of SNPs in high-GC or low-GC regions of the genome (measured as the frequency of guanines and cytosines in a genomic locus), the program OcculterCut v1.1 [39] was used to scan the genomes of A. rabiei strain ArME14 and Leptosphaeria maculans strain v23.1.3 (NCBI accession ASM23037v1 [40]) to assess and compare GC content and distribution. Gene annotations, GC content and frequency of variants were subsequently plotted against the ArME14 genome using CIRCOS [41]. Variants were annotated against the gene models of the ArME14 A. rabiei reference genome using SnpEff v4.3 [38] to identify SNPs with high, moderate, low or modifier (intergenic) impact. For population genetic analyses that require selectively neutral markers (population differentiation, indices of genetic and genotypic diversity and a priori population structure), a subset of variants after the removal of high- and medium-impact sites was used. Identification of clonal lineages An essential step in population genomic analyses of clonal organisms is the identification of clones, often defined as unique multilocus genotypes (MLGs) without inflating the number of clones due to variant-calling errors or missing data [4245]. Our preliminary analyses of the A. rabiei population in Australia demonstrated that many unique MLGs differed at only a small proportion of loci (Fig. S1). We, therefore, used the functions cutoff_predictor and mlg.filter in poppr to collapse all MLGs with genetic distance smaller than the estimated threshold of 0.0045 into the same MLL. Therefore, all MLGs with nine or fewer differences were grouped into the same MLL [4546]. Population differentiation and structure The term population is used here to refer to a group of isolates defined under certain geographical or temporal criteria, e.g. collected in the same year or within the same agroecological zone. Since our dataset included individuals from 2013 to 2020, we first tested for temporal population differentiation to investigate whether individuals collected in different years from the same agroecological zones can be pooled together. The analysis of molecular variance (AMOVA) was conducted using ade4 v1.7–18 [47] in poppr v2.9.3 [46] to test for population structuring within and among different years without making assumptions about the Hardy–Weinberg equilibrium. For this, we used the subset of low-impact and intergenic SNPs (total of 2554 sites) to investigate the partitioning of genetic variation within and among locations, zones and states in each year and among years. The significance of Phi (ϕ) statistic was tested using 999 permutations. Due to the small number of individuals in the 2013 and 2014 populations, the AMOVA was repeated after removing these populations from the dataset. The pairwise Hedrick’s G’’ST [4849], which is a standardized measure of population differentiation calculated by dividing GST for a given marker by the maximum theoretical GST, was estimated using the package mmod v1.3.3 [50]. Jost’s pairwise index of differentiation (D), which measures allelic differentiation between populations [51], was also estimated using mmod. Both G’’ST and Jost’s D may range from 0 (no differentiation) to 1 (complete differentiation). Dendrograms of genetic distance among individuals were produced through the neighbour-joining (NJ) method in poppr based on bitwise genetic distance [52] and also using maximum likelihood algorithm in RAxML v8 [53] using acquisition bias correction for SNP datasets [54]. To investigate temporal and geographical population differentiation, Discriminant Analysis of Principal Component (DAPC) analyses were also conducted in R package adegenet v2.1.5 [55] with a priori populations defined as year of collection and also agroecological zones. Unlike AMOVA, DAPC allowed us to include 2013 and 2014 populations, as this approach is not affected by smaller sample sizes. The optimal number of principal components (PCs) to retain for each analysis was determined using the xvalDapc function, and the DAPC analyses were conducted using adegenet. The existence of an underlying structure without a priori population assignment was assessed using DAPC, which does not assume linkage equilibrium of loci. The reduced SNP dataset including only low-impact and intergenic SNPs was used for this analysis (a total of 2554 sites). The optimal number of clusters was determined using the function find.clusters, and dapc was used to assign individuals into clusters, retaining the number of principal components encompassing 94% of the cumulative variance. Indices of gene and genotypic diversity The lack of significant genetic differentiation among populations from different years allowed us to pool the individuals collected from the same agroecological zones in different years together to calculate indices of genetic diversity using poppr. Allele frequencies for all loci were calculated using vcfTools and visualized in R. The genetic distances between A. rabiei isolates were estimated with the diss.dist function in poppr. Identification of loci associated with isolate aggressiveness To identify loci underlying aggressiveness, the entire dataset including SNPs and INDELs with high, moderate, low and modifier impacts were used (a total of 3283 sites). Linkage disequilibrium (LD) decay plots generated in GAPIT3 [56] showed long-range linkage of loci in A. rabiei (Fig. S2). Therefore, the homoplasy-based genome-wide association (GWAS) analysis implemented in POUTINE [57] was implemented using the entire dataset (total of 3283 sites with high, moderate, low and modifier effect). Ancestral trees required as input for POUTINE were generated in RAxML v8 [53] using acquisition bias correction for SNP datasets [54]. POUTINE can only handle binary phenotypes; therefore, separate analyses were conducted to find genotype associations with high/low, high/medium and low/medium aggressiveness. We also performed a second GWAS analysis in pyseer v1.3.10 [58] using the qualitative 1–9 disease severity scale by Singh [29] as phenotypes, with minimum and maximum allele frequencies of 0.04 and 0.98, respectively, and a whole-genome (elastic net) association model. Pyseer includes strategies to account for clonality and population structure and P value corrections for multiple hypothesis testing, which can lead to false positives in traditional GWAS approaches. As a third approach to detect loci underlying aggressiveness, we used DAPC, which is not affected by linkage of loci. In order to remove the complicating effect of population structure, the association of SNPs with aggressiveness of isolates was assessed within each of the three clusters identified in the initial DAPC analysis using the entire dataset (total of 3283 sites with high, moderate, low and modifier effect). Sub-populations within each cluster were defined as groups of isolates belonging to the same aggressiveness group (high, medium and low) on each of the chickpea varieties tested, and their differentiation was assessed using DAPC. The optimal number of PCs to retain for each analysis was determined using the xvalDapc function, and the DAPC analyses were conducted using adegenet. The function loadingplot was subsequently used to find the SNPs contributing to the separation of isolates belonging to each aggressiveness group. Finally, the pairs of closely related isolates (different at <15 sites) with contrasting aggressiveness (PG0/PG1 versus PG4/PG5) were compared to identify genetic variations potentially associated with aggressiveness. Upon the identification of variations potentially associated with aggressiveness of A. rabiei, the impact and location of the site were further investigated, upstream and downstream coding sequences were identified and predicted proteins were used in NCBI blastp searches to characterize their potential function. The probability of the encoded proteins to be fungal effectors was investigated using Predector v1.2.7 [59] and EffectorP v3.0 [60]. For accurate effector prediction, SignalP v6 [61] and Phobius [62] were implemented for computational evidence of secretion for predicted effectors, as EffectorP recognizes a cytoplasmic signal also in intracellular, non-secreted proteins [60]. For intergenic SNPs, we utilized the Neural Network Promoter Prediction server (https://www.fruitfly.org/seq_tools/nnppHelp.html) to explore whether these variants were located within promoter regions, potentially influencing gene expression. Results Species identification and mating-type screening Mating-type PCR screening of all A. rabiei isolates (n=230) confirmed their identity as A. rabiei and resulted in an amplicon of ~400 bp for all isolates, as expected for MAT1-2. Aggressiveness of A. rabiei isolates Of the 230 isolates assessed, 36 and 24 were aggressive (16%) and highly aggressive (10%), identified as belonging to PG4 and PG5, respectively. A significant increase in the frequency of detection of isolates able to cause severe damage on ‘moderately resistant’ cultivars was observed since 2016 (Fig. 2). In particular, the proportion of isolates that were highly aggressive on PBA HatTrick and Genesis 090, both widely adopted cultivars, had steadily increased. Also, the isolates able to cause severe disease on a major resistance source used within the Australian chickpea breeding program, ICC3996, had emerged. Of most immediate concern for the growers was the proportion of isolates able to cause severe disease on a relatively recently released variety PBA Seamer, increasing from 7.4% in 2018 to 44.4% in 2020 (Fig. 2). Summary of A. rabiei aggressiveness between 2013 and 2020. (a) Frequency of highly aggressive isolates (able to produce a leaf score of at least 7 on >80% and a stem score of at least 7 on >10% of the seedlings, see the ‘Assessment of host reaction and assignment of isolate aggressiveness’ in the ‘Methods’ section for details) on each chickpea host genotype at each year. (b) Distribution of isolates by pathogenicity group classification (cumulative disease response to the set of host genotypes; see the ‘Assessment of host reaction and assignment of isolate aggressiveness’ in the ‘Methods’ section for details) in each year. These figures were retrieved from the Ascochyta dashboard on 30 June 2022. Updated information from recent years can be found at https://bit.ly/asco-dashboard. Quality assessment and filtering Illumina sequencing of the 230 A. rabiei isolates resulted in a minimum of 35× coverage per isolate with an average GC% nucleotide content of 48–50% . Filtering of the VCF file output from Freebayes for high-quality and depth variants resulted in a total of 13 259 variants for 230 A. rabiei isolates. Further filtering for the quality, depth, minor allele frequency of 0.01 and removal of loci that gave contradictory results for internal replicates resulted in a total of 3283 high-quality variants sites, including 2723 SNPs (the term SNP here is used to refer to those SNVs that passed the polymorphism threshold of 1%) and 560 INDELs. SNP annotation in SnpEff assigned 13, 156 and 134 SNPs in the final dataset as high, moderate and low impact (present in the coding region of a gene). The remaining 2420 SNPs were identified as modifier (intergenic) loci. For INDELs, 7, 49, 11 and 493 sites were identified as high, moderate, low and modifier impact. High- and moderate-impact variants are predicted to have a disruptive effect on protein function (truncation, loss of function or nonsense-mediated decay) or change protein effectiveness, while low-impact variants are assumed to be harmless and unlikely to change protein function. Modifier (or intergenic) variants, on the other hand, reside in non-coding regions of the genome, where predictions are difficult or there is no evidence of impact [38]. For population genetic analyses that require selectively neutral markers, SNPs with high and moderate impact were removed from the dataset, and only biallelic SNPs with low and modifier impacts were used (total of 2554 sites). The analysis using OcculterCut revealed that 22.5% of A. rabiei genome consisted of AT-rich regions (measured as the frequency of adenines and thymines in a genomic locus, see Fig. 3), which also harboured the majority of the SNPs. The position of SNPs on the A. rabiei genome and in relation to GC per cent and gene content is shown in Fig. S3. The GC content distribution of A. rabiei strain ArME14 in comparison with another plant pathogenic fungus, L. maculans strain v23.1.3. Vertical blue lines show the GC cut-off points selected by OcculterCut [37] to classify genome segments into AT-rich and GC-balanced regions. The percentage values shown on the left and right sides of the vertical blue lines indicate the percentage of the genome classified as AT-rich and GC-balanced, respectively. Population differentiation No significant differentiation was detected among A. rabiei collected in different years after removing the smaller 2013 and 2014 populations (Tables2 3). Low but statistically significant genetic differentiation was detected between populations from different states (14%) and agroecological zones (21%), and most of the genetic variation (66%) was found among individuals within states and agroecological zones (P=0.001). This indicated low but significant structuring of the population based on isolate geographical location, while no significant temporal differentiation was detected among populations collected from 2015 to 2020 using AMOVA. Adding 2013 and 2014 populations to the analyses did not change the results of AMOVA analyses (Tables S2 and S3). Table 2.Summary of the AMOVA results for 223 A. rabiei isolates collected in 2015, 2016, 2017 and 2020 from different states and locations in AustraliaSourced.f.*SS†MS‡VarianceProportion of variation (%)P value Among years 31471.32490.440.4500.423 Among states within years 107834.44783.4431.7013.50.001 Among locations within states 3412 111.72367.0246.5720.50.001 Within samples 17525 207.65144.87144.8766.00.001 Total 22246 625.130211.93222.69100.0–*Degrees of freedom.†Sum of squared observations.‡Mean of squared observations. Summary of the AMOVA results for 223 A. rabiei isolates collected in 2015, 2016, 2017 and 2020 from different states and locations in Australia Degrees of freedom. Sum of squared observations. Mean of squared observations. Table 3.Summary of the AMOVA results for 223 A. rabiei isolates collected in 2015, 2016, 2017 and 2020 from different agroecological zones and locations in AustraliaSourced.f.*SS†MS‡VarianceProportion of variation (%)P value Among years 31471.32490.444.652.00.185 Among zones within years 1913 297.69738.7641.4618.00.001 Among locations within zones 256.648.48265.9330.1513.50.001 Within samples 17525 207.65144.87144.8766.50.001 Total 22246 625.13211.93221.14100.0–*Degrees of freedom.†Sum of squared observations.‡Mean of squared observations. Summary of the AMOVA results for 223 A. rabiei isolates collected in 2015, 2016, 2017 and 2020 from different agroecological zones and locations in Australia Degrees of freedom. Sum of squared observations. Mean of squared observations. Jost’s and Hedrick’s G’’ST indices of pairwise index of differentiation (D) were also low between populations from different years (Table 4). On the other hand, DAPC analysis using year of collection as a priori population assignment, which is not affected by population size and allowed us to also include populations from 2013 and 2014 in the analyses, separated the majority of the 2013 isolates from the rest of the population (Fig. 4a). Structuring of the A. rabiei population based on geographical location (agroecological zones) was confirmed by the results from the DAPC analysis (Fig. 4b). Table 4.Pairwise population differentiation (Jost’s D) between A. rabiei populations collected in different years in Australia shown at the bottom of the table and pairwise population differentiation (Hedrick’s G’’ST) between A. rabiei populations collected in different years in Australia at the top of the table in parentheses 2015 2016 2017 2020 2015 0(0.142)(0.206)(0.232) 2016 0.0100(0.039)(0.047) 2017 0.0150.0020(0.078) 2020 0.0200.0030.0060 Pairwise population differentiation (Jost’s D) between A. rabiei populations collected in different years in Australia shown at the bottom of the table and pairwise population differentiation (Hedrick’s G’’ST) between A. rabiei populations collected in different years in Australia at the top of the table in parentheses (a) DAPC for A. rabiei isolates collected in Australia from 2013 to 2020, using year of collection as a priori populations. The optimal number of PCs to retain for each analysis was determined by xvalDapc function [55] to be 32, encompassing 88% of the cumulative variance. The ellipses represent the maximum area spanned by 95% of the data in a population by year of collection. (b) DAPC for A. rabiei isolates collected in Australia from 2013 to 2020, using agroecological zones (Table 1) as a priori populations. The optimal number of PCs to retain for each analysis was determined by xvalDapc function [55] to be 58, encompassing 98% of the cumulative variance. The ellipses represent the maximum area spanned by 95% of the data in an agroecological zones. Genetic diversity The A. rabiei population consisted of 167 multilocus lineages (MLLs) and a genotypic diversity of 0.99 (Simpson’s complement λ=0.99) [63], high evenness (E5=0.85) and low clonality (clonal fraction=26%). Despite the high genotypic diversity, gene diversity (allelic diversity) of the population was very low (Hexp ≈ 0) due to the low level of polymorphism in the majority of loci (median allele frequency=0.017). Therefore, despite the identification of 167 unique MLLs, these clonal lineages differed at only a small proportion of loci. The maximum number of allelic differences between genotypes was 820 SNPs, and the percentage of SNP differences between genotypes (the ratio of the number of observed differences by the number of possible differences) was 14% (Fig. S1). Population structure Population structure analysis with no a priori population assignment through DAPC analysis detected three distinct clusters in the population (Fig. 1b). Although the Bayesian information criterion (BIC) consistently decreased with the number of clusters, K=3 was selected to best explain the population structure since the BIC value decreased only incrementally when K >3. The detected three clusters consisted of 43 (cluster 1), 66 (cluster 2) and 121 (cluster 3) isolates, which did not correspond to sampling locations, agroecological zones or states. The first two discriminant functions explained 96% of the total conserved variance. The NJ dendrogram of the isolates based on bitwise genetic distance using 2554 putatively neutral SNPs was generally in agreement with the DAPC results (Fig. 5). The isolates belonging to DAPC cluster 2 were further divided into two sub-clusters 2a and 2b in the NJ dendrogram (Fig. 5). Cluster 2a included the highest frequency (53%) of aggressive isolates [PG4 and PG5, followed by cluster 3 (26%) and cluster 2b (19%)]. Cluster 1 included the lowest frequency of aggressive isolates with only one isolate belonging to PG5 and three isolates belonging to PG4. Interestingly, all isolates in cluster 2a were from Western Australia (WA), except for a single, highly aggressive isolate (AR0304) from the southern growing regions in South Australia and Victoria (SA/Vic1). Repeating the NJ analysis using the entire dataset including moderate- and high-impact loci (a total of 3283 loci) did not change the results (Fig. S4). The dendrogram generated via maximum-likelihood algorithm in RAxML resulted in a similar topology (data not shown). The NJ dendrogram of 130 A. rabiei isolates sequenced in this study based on bitwise genetic distance using 2554 putatively neutral SNPs. Tip colours correspond to the aggressiveness (pathogenicity group) of the isolates as indicated in the legend. Branches coloured green, orange and purple denote isolates identified in clusters 1, 2 and 3 in the DAPC analysis shown in Fig. 1. Identification of SNPs potentially associated with A. rabiei aggressiveness GWAS analysis using POUTINE did not find any variants significantly associated with the aggressiveness of A. rabiei on different host varieties. While multiple mutations were identified to have a significant pointwise P value (<0.05), none survived correction for multiple hypothesis testing (family-wise P value>0.05). GWAS analyses using pyseer, on the other hand, identified 66 variants to be significantly associated with the aggressiveness of A. rabiei on different host varieties (Fig. 6, Table S5). An SNP at locus ctg_08_1872963 was identified to be significantly associated with A. rabiei aggressiveness on all four chickpea cultivars, and SNPs at loci ctg_19_1154999 and ctg_19_1155004 were associated with aggressiveness on three cultivars (Genesis090, PBA HatTrick and PBA Seamer). Of the remaining variations, 17 were associated with two chickpea cultivars, while 34, 6, 5, and 1 variation were associated with PBA Seamer, PBA HatTrick, ICC3996 and Genesis090 only. The majority of detected variations were located in intergenic regions, with only six variations located on coding regions with low to moderate impact (Table S5). Manhattan plot showing variations with statistically significant association (the significance threshold at P value <0.05, indicated by the red horizontal dotted line) with aggressiveness of A. rabiei in Australia isolates based on pyseer [58] results. The DAPC analysis for each of the three A. rabiei clusters using isolates aggressiveness towards each chickpea variety (low, moderate and high) as pre-defined populations showed various degrees of separation. A total of 151 variations were identified to be associated with aggressiveness on different chickpea cultivars (loading >0.005), 22 of which were located in coding regions with low to moderate impact, and the rest were intergenic variations (Fig. 7, Table S6). Of these, five SNPs (ctg05_2328653, ctg14_1473121, ctg14_1473129 and ctg15_1503638, ctg15_1503666) were found to be linked to aggressiveness of isolates on three chickpea cultivars. Twenty-five SNPs were linked to aggressiveness on two chickpea cultivars (Table S6). SNP markers potentially associated with aggressiveness of A. rabiei isolates on four different chickpea cultivars, namely ICC3996, PBA Seamer, Genesis090 and PBA HatTrick identified using DAPC analyses of three subpopulations (clusters) detected in Australia. The red horizontal line delineates the loading threshold of 0.01. Sixteen variations were identified to be associated with A. rabiei aggressiveness in both DAPC and pyseer analyses (Table 5). Of the remaining variations, although the same variation was not a hit in both analyses, the coding regions associated with the variation were implicated in both DAPC and pyseer analyses. For example, ctg_08_1872963 variation, which was identified in pyseer to be associated with A. rabiei aggressiveness on all chickpea cultivars, was not a hit in the DAPC analyses, but other closely linked variations with the same predicted impact were identified in the DAPC analysis (intergenic, modifier variations downstream EKO05_005581 coding region). Table 5.Genetic variations predicted to be associated with aggressiveness of A. rabiei isolates on different chickpea cultivars in Australia, detected using both GWAS (pyseer) and DAPC analysesVariation positionP valueVariation type, impact and location on the genomePutative protein functionPredector score (EffectorP 3.0 score)Repeat familyGenesis090ICC3996PBA HatTrickPBA seamerctg_01_1654657–––0.0273Intergenic_region, modifier, EKO05_000484-EKO05_000485EKO05_000484: hypothetical protein with high similarity to fungal GTPases, EKO05_000485: hypothetical DNA binding protein with a transmembrane domainNon-effectorsLTR/Copiactg_01_1665154–––0.0273Same as ctg01_1654657ctg_02_2690139–––0.0273Missense_variant, moderate impact, EKO05_001768Hypothetical protein with similarity to fungal transcription factors with a zinc-finger domainNon-effector–ctg_02_364191–––0.0273Intergenic_region, modifier, EKO05_001071-EKO05_001072EKO05_001071: hypothetical HET domain containing protein with a transmembrane domain, EKO05_001072: hypothetical phosphotransferaseNon-effectorsLTR/Gypsyctg_03_1415178–––0.0273Intergenic_region, modifier, EKO05_002061-EKO05_002062EKO05_002061: hypothetical protein of unknown function, EKO05_002062: hypothetical apoplastic effectorEKO05_002061: non-effector, EKO05_002062 : 0.676 (0.709, apoplastic effector)LTR/Gypsyctg_08_448981–––0.0364Intergenic_region, modifier, EKO05_005280-EKO05_005281EKO05_005280-EKO05: hypothetical apoplastic effector, EKO05_005281: hypothetical protein of unknown functionEKO05_005280: −2.911 (0.509, apoplastic effector), EKO05_005281: non-effectorLTR/Gypsyctg_08_98214–––0.0273Intergenic_region, modifier, EKO05_005201-EKO05_005202EKO05_005201: hypothetical apoplastic effector, EKO05_005202: hypothetical protein of unknown functionEKO05_005201: −1.476 (0.654, apoplastic effector), EKO05_005202: −3.378 (0.675, cytoplasmic effector but no signal peptide detected)LTR/Gypsyctg_10_10643––0.008950.0374Synonymous_variant, low impact, EKO05_006140;on a potential promoter region with a prediction score of 0.9Hypothetical protein of unknown function−2.614 (0.522, cytoplasmic effector but no signal peptide detected)DNA/Kolobok-Hctg_10_157955–––0.0364Intergenic_region, modifier, EKO05_006148-EKO05_006149EKO05_006148: hypothetical protein of unknown function, EKO05_006149: hypothetical oxidooreductaseNon-effectorsDNA/TcMar-Fot1ctg_10_157964–––0.0438Same as ctg_10_157955ctg_10_158053–––0.0364Same as ctg_10_157955ctg_12_1234964–––0.0273Intergenic_region, modifier, EKO05_007471-EKO05_007472EKO05_007471: hypothetical protein of unknown function, EKO05_007472: hypothetical cytoplasmic effectorEKO05_007471: non-effector, EKO05_007472: −1.071 (0.535, cytoplasmic effector, signal peptide detected)LTR/Gypsyctg_12_1234984–––0.0273Same as ctg_12_1234964ctg_15_15036960.0199–––Downstream_gene_variant, modifier, EKO05_008905EKO05_008905: hypothetical DNA helicase with a transmembrane domainNon-effectorsUnknown familyctg_19_11549990.00638–0.002930.000122Intergenic_region, modifier, EKO05_010492-EKO05_010493EKO05_010492: hypothetical apoplastic effector, EKO05_010493: amino acid transporter with multiple transmembrane domainsEKO05_010492 : 2.244 (0.895, apoplastic effector), EKO05_010493: non-effectorLTR/Gypsyctg_19_11550040.00638–0.002930.000122Same as ctg_19_1154999*Predector score is provided with EffectorP 3.0 score in parentheses. Non-effector indicates a negative Predector score and >0.5 non-effector score from EffectorP 3.0. Genetic variations predicted to be associated with aggressiveness of A. rabiei isolates on different chickpea cultivars in Australia, detected using both GWAS (pyseer) and DAPC analyses Predector score is provided with EffectorP 3.0 score in parentheses. Non-effector indicates a negative Predector score and >0.5 non-effector score from EffectorP 3.0. The phylogenetic analysis showed that aggressive isolates developed independently multiple times across the tree (Fig. 5). The comparison of pairs of closely related individual isolates (different at <15 loci) with contrasting aggressiveness (PG0/PG1 versus PG4/PG5) identified genetic variations, which were also detected through pyseer and DAPC analyses to be potentially associated with A. rabiei aggressiveness (Table S7). Discussion Population genomic analyses of Australian A. rabiei isolates using genome-wide SNPs confirmed its clonality in Australia, in line with results from previous studies that reported the lack of sexual reproduction [613 14] as opposed to elsewhere in the world, where A. rabiei is known be sexually recombining and genotypically diverse [1517]. The three main clonal lineages detected in the Australian A. rabiei population are not evenly distributed in all agroecological zones. Isolates from cluster 3 were found in all regions, while cluster 1 isolates are absent from WA and northern NSW and Qld zones, and cluster 2 isolates are absent from northern NSW and Qld zones. The presence of all three lineages in the southern agroecological zones indicated higher genotypic diversity of A. rabiei in these regions and is in line with previous findings that South Australia is most likely where A. rabiei was first introduced to Australia [13]. This also emphasizes the continuous need for biosecurity measures to inhibit long-distance dispersal of the pathogen within Australia, which may result in fundamental changes to A. rabiei populations in WA and Qld. The genotyped population was further phenotyped for aggressiveness on different chickpea cultivars to explore genetic variations potentially associated with the aggressiveness of A. rabiei in Australia. This is the first study to reveal the complex nature of aggressiveness in the Australian A. rabiei population, where aggressive isolates have evolved independently several times within different clonal lineages and as a result of multiple small-scale genetic variations. Classic genome-wide association studies, which are commonly used for the identification of loci associated with target phenotypic traits in fungal plant pathogens with frequent cycles of sexual reproduction [6466], are only relevant in the presence of recombination [67]. Such methods are not appropriate for identifying associations of phenotype and genotypes in strongly clonal populations with long-range LD and population stratification. While non-sexual recombination (for example, parasexuality) may still occur in asexual organisms, significant LD of loci and high clonality of A. rabiei in this study and previous research point to the lack of recombination in the Australian A. rabiei population. In such situations, homoplasy-counting GWAS methods are more appropriate because these methods make use of homoplasic alleles (identical-by-state but not identical-by-descent), which arise repeatedly and independently in different genetic backgrounds and, thus, are not in LD [56]. To detect loci that may be associated with aggressiveness of A. rabiei in Australia, we made use of DAPC analysis, which does not require assumptions of linkage equilibrium of loci [55], as well as pyseer, which accounts for confounding effects of population structure and clonality [58] and POUTINE [56], a homoplasy-counting GWAS method. DAPC resulted in the identification of 151 variations contributing to the separation of A. rabiei isolates based on their aggressiveness on different cultivars and within different clonal lineages. Only 22 of these were located on coding sequences: 14 causing missense mutations and 8 resulting in synonymous substitutions (Table S6). Pyseer analysis detected 66 variations underlying aggressiveness of A. rabiei isolates, only 6 of which were located on coding regions (2 synonymous mutations and 4 missense mutations, Table S5). Of these, 16 variations were detected in both analyses, including ctg_02_2690139, which resulted in a missense variation on a gene coding for a hypothetical transcription factor with a zinc-finger domain, a known regulator of pathogenicity in fungal plant pathogens [6870]. Several variations detected using DAPC analyses and pairwise comparison of genetically close isolates with contrasting aggressiveness (ctg05_2388561, ctg14_6054, ctg14_1527717, ctg15_1503638 and ctg20_6490) resulted in missense mutations in genes coding for hypothetical DNA helicases. Several other variations were found in association with synonymous changes in DNA helicase genes or upstream and downstream of putative DNA helicase genes. All these variations were located in telomeric regions. DNA helicases are nucleic acid-dependent ATPases with the ability to separate DNA strands, which is an important step in genome replication, expression and repair [71]. Therefore, helicases have diverse functions in the cell, and their substrates substantially vary, as they facilitate almost all transactions in nucleic acid metabolism. In fungi, helicases have been associated with post-transcriptional gene silencing [72]. Telomere-linked helicases have been detected in subtelomeric DNA of filamentous fungi such as Metarhizium anisopliae [73], Ustilago maydis [74] and Magnaporthe oryzae [7576], and their function is currently unclear, although it may be related to the maintenance of genome stability [76]. An RNA helicase gene in M. oryzae encodes for the MoDHX35 protein, which was reported to have a role in appressorium formation and virulence of this pathogen on rice [76, 77]. In the mycoparasitice fungus, Coniothyrium minitans, the disruption of a DNA helicase gene resulted in its altered morphology, reduced growth rate and loss of mycoparasitism towards Sclerotinia sclerotiorum [78]. Telomeric regions in several fungi have been reported to be highly dynamic and accommodate genes that confer an adaptive advantage in evading host recognition [7980]. Therefore, it is plausible that the identified variations in association with DNA helicases in sub-telomeric regions may have a role in transcription and epigenetic changes associated with aggressiveness in A. rabiei. Many of the SNPs identified in the current study occur on genes encoding hypothetical proteins of unknown function with no predicted signal peptide or transmembrane domain. For example, ctg07_899731 and ctg21_487980 identified by pyseer and ctg13_5391 and ctg13_5645 identified through DAPC analyses and pairwise comparison cause missense variations on genes coding for hypothetical proteins of unknown function. These highlight the limitation of the current reference genome of A. rabiei, which still lacks accurate functional annotation for many hypothetical predicted genes. Further studies are required to understand the potential role of these loci in the aggressiveness of A. rabiei on chickpea. Variations ctg07_1292465 and ctg20_254353 were found to cause missense mutations in genes coding for a hypothetical alkaline phosphatase and a hypothetical ubiquitin transferase, both of which have been associated with fungal pathogenesis on various pathosystems [8182]. Several variations, however, were identified in intergenic regions in the vicinity of coding sequences that code for hypothetical proteins with potential function in pathogenicity and aggressiveness. These include genes encoding a secreted carbohydrate-binding trehalase (EKO05_010756), a secreted protein with similarity to glycosylphosphatidylinositol (GPI)-anchored, serine–threonine-rich proteins (EKO05_007578) and multiple hypothetical proteins with transmembrane domains and high similarity to sugar, oligopeptide and copper transporters. Trehalases have an important role in the hydrolysis of external trehalose in fungi and bacteria, where they play an important role in pathogenicity, virulence and stress resistance [74,8385]. Some trehalases may have other cellular functions, e.g. a trehalase-encoding gene was primarily expressed in spores of Plasmodiophora brassicae, believed to release glucose as a source of energy during the germination process [86]. GPI-anchored, serine–threonine-rich proteins in Candida albicans have been reported as key effectors of fungal adherence to host cells [87]. The role of transmembrane proteins in cellular transport, metabolism, signal transduction, conidial germination and aggressiveness of fungal plant pathogens is well documented [75,8891]. Several variations associated with A. rabiei aggressiveness on chickpea varieties were found in the proximity of loci coding for predicted apoplastic or cytoplasmic effectors on the reference genome (strain ArME14). Effectors are secreted proteins that facilitate infection through suppression of plant defence responses and alteration of host cell structure and function [92]. Apoplastic effectors are secreted into the host extracellular space and can function in the apoplast or may be bound to the fungal cell wall [9394]. On the other hand, cytoplasmic effectors are secreted into the cytoplasm and may subsequently target specific plant cell compartments [60]. One of the genes predicted to encode for an apoplastic effector (EKO05_005328) shows high similarity to fungal genes encoding rapid alkalinization factors (RALFs). The members of the RALF family are a family of conserved plant regulatory peptides, homologues of which have also been detected in plant pathogenic fungi with potential role in pathogenicity. For example, Fusarium oxysporum mutant strains lacking a functional RALF showed reduced virulence in tomato plants [95], and a RALF peptide has been suggested to be a minor contributor to Fusarium graminearum virulence in wheat [96]. The presence/absence of these effectors in different A. rabiei isolates and the potential for differential expression of these effectors in isolates with different levels of aggressiveness should be further investigated. The investigation of the potential association of SNPs with recently functionally characterized A. rabiei effectors detected one modifier SNP located upstream of the gene encoding ArCRZ1 protein (Table S8), which is a putative transcriptional regulator of virulence factors in A. rabiei [97]. Many genetic variations detected here were also located within transposable elements (TEs) and repetitive regions of the genome. Most of these repeat regions consisted of TEs, including the only SNP that was commonly associated with aggressiveness across all host chickpea cultivars, which occurred within a retrotransposon of the LTR/Gypsy family, known to be involved in rapid increases in fungal pathogenicity [98]. Transposable elements are self-replicating segments in the genome, which are the target of epigenomic regulation that is also known to affect nearby genes. In other words, TEs have the ability to also alter gene regulation [99] and have been implicated in differential expression of effector loci and pathogenicity-related genes in plant pathogenic fungi [100101]. Our study revealed that 22.5% of A. rabiei genome consists of AT-rich regions (secondary peak in Fig. 3), which are known to harbour repetitive regions and TEs. This is similar to the reported value of 21% reported for an earlier version of A. rabiei genome [31]. This phenomenon in fungi is referred to as the ‘two-speed genome’ [102], where effectors with an important role in the virulence and aggressiveness of plant pathogens lie within rapidly evolving, AT-rich regions with highly repetitive content and increased TE activity. This will allow the effector loci within the fast-evolving, repetitive genome compartments to evolve through duplication, modification and inactivation or be lost entirely in a short period of time due to the activity of TEs. Therefore, although genetic variations detected in the repetitive regions of the genome were annotated as intergenic, these could potentially be located on effector loci, which have not been annotated due to their highly divergent sequences and low homology to known effectors. High levels of the presence/absence polymorphisms of effectors have been reported within fungal plant pathogens such as M. oryzae [103] and Blumeria graminis [86]. Although the percentage of AT-rich repetitive regions in A. rabiei is considerably lower than many other aggressive plant pathogens such as L. maculans (shown in the secondary peaks left of the vertical line in Fig. 3), our results show that a high rate of variation in these regions may be a major driving force underlying the rapid evolution of aggressiveness in the A. rabiei population in Australia. A commonly used approach for investigating genetic loci underlying adaptive mechanisms, which has been successfully applied to plant pathogens, is investigating signatures of positive selection. Two commonly used approaches to detect signatures of selection are the dN/dS ratio test and selective sweep scans [104]. The dN/dS ratio test is more appropriate for macroevolutionary scenarios spanning speciation events [105], while sweep scans can be used to investigate the recent evolution of populations undergoing strong positive selection similar to the situation for A. rabiei in Australia. However, the detection of selection sweeps is reliant on the presence of sexual reproduction in pathogen populations, where loci undergoing selective sweeps can be differentiated from neighbouring loci because of a significant reduction in sequence diversity, on which meiotic recombination has not yet had a chance to act to break the linkage of loci. Such analyses are, therefore, not appropriate for a highly clonal, asexually reproducing pathogen [106] like A. rabiei in Australia, with very low levels of genome wide polymorphism and a genome that is effectively entirely linked (Fig. S2). This study presents the first step towards understanding the molecular mechanisms of aggressiveness in A. rabiei populations in Australia. Detection of multiple SNPs with an incremental contribution towards aggressiveness of A. rabiei isolates is in line with the phenotypic observation demonstrating a quantitative variation in the pathogenicity of A. rabiei isolates on different chickpea varieties. Differences in the aggressiveness of pathogens have previously been suggested to be conditioned by minor gene‐for‐minor gene interactions and variation in aggressiveness traits defined as quantitative life‐history traits, such as infection efficiency, latent period, sporulation rate and lesion size [107]. The limited number of small variant (SNPs) differences between pairs of isolates with contrasting aggressiveness suggest that other forms of genetic modifications, including large structural variants such as accessory chromosomes, gene presence/absence, copy number variations and large indels and epigenetic modifications, may have an important role determining A. rabiei aggressiveness on chickpea. The short-read sequencing approach performed here proved cost-efficient to generate high-coverage, high-quality, genome-wide SNP variants on a population scale on a small genome such as A. rabiei and was effective in identifying underlying population structure using a larger number of markers ever used. However, short-read data could not be used to identify larger structural and epigenetic variations. Therefore, our future efforts will focus on long-read sequencing of genetically close isolates with contrasting aggressiveness as well as epigenomic investigations to further unravel the underlying mechanisms of aggressiveness in A. rabiei. Identifying the genetic regions contributing to these traits will help with the rapid and reliable detection of highly adapted and aggressive isolates. supplementary material