Widespread genetic heterogeneity and genotypic grouping associated with fungicide resistance among barley spot form net blotch isolates in Australia

Abstract Spot form net blotch, caused by Pyrenophora teres f. maculata, is a major foliar disease of barley worldwide. Knowledge of the pathogen's genetic diversity and population structure is critical for a better understanding of inherent evolutionary capacity and for the development of sustainable disease management strategies. Genome-wide, single nucleotide polymorphism data of 254 Australian isolates revealed genotypic diversity and an absence of population structure, either between states, or between fields and cultivars in different agro-ecological zones. This indicates there is little geographical isolation or cultivar directional selection and that the pathogen is highly mobile across the continent. However, two cryptic genotypic groups were found only in Western Australia, predominantly associated with genes involved in fungicide resistance. The findings in this study are discussed in the context of current cultivar resistance and the pathogen's adaptive potential.


Introduction
Spot form net blotch (SFNB) is a major foliar disease of barley worldwide (Liu et al. 2011), caused by the ascomycete fungus Pyrenophora teres f. maculata (Ptm) (Smedegård-Petersen 1971). The pathogen is morphologically similar, yet phylogenetically distinct from Pyrenophora teres f. teres (Ptt), the causal agent of net form net blotch (NFNB) disease. The two diseases are distinguished by their physiological leaf symptoms, with SFNB characterized by round brown lesions surrounded by a yellow chlorotic halo and NFNB characterized by dark brown, net-like necrotic lesions striate along barley leaf veins. Despite occurring on the same host and their similar morphological characteristics, the two diseases are treated separately as they interact with different host resistance and susceptibility genes (Clare et al. 2020) and, although natural hybridization is possible, this is rare (Poudel et al. 2017). In Australia, SFNB is of economic importance, with yield losses of up to 20% and an increase of up to 18% in undersized grain (McLean et al. 2022).
Fungal pathogen evolution in plants is broadly governed by effectors, small rapidly evolving secreted proteins or molecules, which interact with the host in different ways dependent on pathogen lifestyle (Plissonneau et al. 2017). For example, biotrophs secrete effectors that suppress host defenses and subvert metabolism to allow proliferation (Dodds et al. 2009). Necrotrophs, by contrast, secrete effectors that promote cell death (Faris and Friesen 2020). Host recognition and effector diversification drive the interaction between fungal pathogens and plants, with dominant host resistance against biotrophs leading to the loss or alteration of the corresponding virulence (effector) gene. In necrotrophs such as Ptm, there is a predominantly opposite genetic relationship, known as the inverse gene-for-gene model, where susceptibility is dominant as genes encoding host target products are lost or altered (Peters Haugrud et al. 2019;Muria-Gonzalez et al. 2023).
The genetic structure and diversity of Ptm has been previously assessed within Australia using both simple sequence repeats (SSRs, Bogacki et al. 2010;McLean et al. 2010) and amplified length polymorphism (AFLP) markers (Serenius et al. 2007;Lehmensiek et al. 2010;McLean et al. 2014). These studies analyzed modest numbers of both isolates (31-60) and polymorphic markers (15-109) and found low genetic differentiation between different regions and fields and no correlation with geographic origin, although both Serenius et al. (2007) and McLean et al. (2010) found differences in the levels of genetic diversity between regions.
Globally, genetic diversity studies of Ptm have been performed in Europe (Rau et al. 2003;Bakonyi and Justesen 2007), North America (Akhavan et al. 2016), Algeria (Ahmed Lhadj et al. 2022), Iran (Vasighzadeh et al. 2021), the Republic of South Africa (Campbell et al. 2002), and Turkey (Oğuz et al. 2019). In common with previous Australian studies, these employed older genetic marker technologies including SSRs, inter-simple sequence repeats, and AFLPs and generally found significant levels of sexual recombination and low levels of clonality. The studies in both Sardinia and Canada found no significant genetic differentiation between populations. In Iran, however, using the same SSRs as those in the Canadian study, higher genetic variation was apparent together with significant evidence for regional population structure, while Lehmensiek et al. (2010) were able to distinguish isolates from South Africa and Australia.
Although microsatellites and AFLP DNA markers have been commonly used for comparing individuals with high levels of genetic diversity, they usually lack high marker numbers (in the case of microsatellites) and precise genomic context and even genomic distribution (in the case of AFLPs) available with genotype-bysequencing (GBS) techniques, where large, single nucleotide polymorphism (SNP) datasets are produced. GBS techniques have been shown to provide clearer detection of finer scale genetic structuring compared to microsatellite data (Sunde et al. 2020). Furthermore, hundreds or thousands of SNP markers generated throughout the target genome produce higher resolution data on smaller sample sizes compared to SSRs (Jeffries et al. 2016). This allows better comparisons between both strongly and weakly diverged populations (Andrews et al. 2016), as well as better inferences on population structure (Bruneaux et al. 2013). One popular method of generating SNP markers is by Diversity Arrays Technology (DArT), a highly parallel genome-wide approach based on a restriction enzyme complexity reduction step that selects for non-repetitive (coding) DNA (Sansaloni et al. 2011). The DArT system produces reproducible silicoDArT (presence-absence markers) and DArTseq (SNP) markers from genomic DNA extracts. This technology has been previously implemented for genetic mapping and genotypic analyses of fungal species including P. teres (Syme et al. 2018;Poudel et al. 2019).
We hypothesized that the composition of the Ptm population across Australia is likely to be influenced by geographic isolation and the cultivars (cvs) being grown, and that low sample numbers together with few or potentially clustered genetic markers in previous studies may have excluded genomic regions contributing to population structure. To assess this, and to provide greater resolution of the genetic diversity and the distribution of Ptm genotypes in Australia, we analyzed isolates based on DArTseq SNP markers. These were sampled at two levels: on the one hand isolates collected from across the major barley-growing regions and, on the other hand, isolates collected from six fields at three sites in Western Australia. The isolates were collected between 2017 and 2020, with the majority of Western Australian isolates collected in 2020. Their genetic diversity was then assessed, and the relationships between isolates were compared by Bayesian, multivariate, and geographic distance-based approaches.

Field-scale Ptm isolate collections
Isolates were sampled from six fields across three agricultural zones (Agzones) of Western Australia (WA) for inter-field analyses (Supplementary Table 2, Fig. 1). Agzones 2, 3, and 6 were chosen due to their status as high rainfall zones (>450 mm per year) suited to barley growth, whilst spanning the majority of the barleygrowing area of WA. Two fields per Agzone were sampled, one of barley cv Spartacus CL and the other of cv RGT Planet. In each field, samples were obtained from a 60 m × 100 m area divided into 100 m 2 sections. Three leaves displaying SFNB symptoms were collected from each section and stored in paper envelopes.

Inter-regional collections of Ptm isolates and related Pyrenophora sp.
To investigate national Ptm diversity, a total of 87 Ptm isolates were provided by the Centre of Crop and Disease Management, the South Australian Research and Development Institute and Agriculture Victoria (AgVic, Fig. 1). The isolates were from locations dispersed across the majority of major barley-growing regions of WA, South Australia (SA), Victoria (Vic), New South Wales (NSW), and Queensland (Qld). Included in the regional Ptm collections were four randomly chosen Ptt isolates from the same institutions.

Fungal isolation
To obtain single-spored isolates, leaf samples were dried at room temperature for 2 weeks, then surface sterilized (30 seconds in 15% ethanol, followed by 30 seconds in 5% ethanol and 1% bleach, and rinsed two times for 30 seconds in sterile water). Net blotchlike lesions were excised and placed on a petri dish containing sterile paper towel wetted with sterile water, sealed with Parafilm (Bemis Inc., Neenah, WI, USA) and incubated at 18°C with a 14-hour photoperiod for up to 7 days. Leaves were inspected for conidia formation daily under a binocular microscope from the third day post plating. Cultures where no conidia were produced after 7 days were placed under near UV light for 18 hours at room temperature, followed by 24 hours in the dark at 15°C to stimulate sporulation. Mature conidia were collected with a sterile acupuncture needle and transferred to V8-PDA agar plates [150-ml/l V8 juice (Campbell's Soups Australia, Lemnos, VIC, Australia), 10-g/l agar (Oxoid Ltd., Basingstoke, UK), 10-g/l Difco potato dextrose agar (Becton Dickinson, Sparks, MD, USA), and 3-g/l CaCO3]. Plates were incubated at room temperature for 5 days before 4-mm 2 plugs were cut from each colony, then air-dried in a biosafety cabinet overnight before storage at −80°C.

DNA extraction and identification of P. teres mating type loci
Cultures were grown for 5 days in Fries 2 liquid medium (Fries 1938). The mycelia were lyophilized and genomic DNA isolated using a Promega Wizard Genomic DNA Purification Kit (Fitchburg, Wisconsin, USA) according to the "Isolating genomic DNA from plant tissue" protocol. DNA concentration and quality were measured with a NanoDrop spectrophotometer (Thermo Fisher Scientific, MA, USA). MAT (mating-type) locus mating type and formae were determined by the polymerase chain reaction (PCR) method according to Lu et al. (2010).

Genotype-by-sequencing and DArTseq data filtering
Pyrenophora teres isolates were genotyped using DArTSeq whole genome sequencing at Diversity Arrays Technology Pty Ltd (Canberra, Australia). DArT is a highly parallel, genome-wide genotyping technology developed by Jaccoud et al. (2001) combined with next-generation sequencing (Sansaloni et al. 2011) and proprietary analytical pipelines (Ren et al. 2015). DArTSeq genome complexity reduction was performed with PstI and MseI restriction enzymes, followed by next-generation sequencing using a HiSeq2000 DNA sequencing platform (Illumina, USA). In total, 5110 DArTSeq (SNP markers) and 6321 silicoDArT (presence/absence markers) were produced. DArT genotype data is provided in Supplementary Table 1. SNP markers were converted into the Genalex file format (Peakall and Smouse 2012) and imported into RStudio 4.2.0 (RStudio, Boston, MA, USA). Markers and isolates with >10% missing data were removed. Phylogenetically uninformative loci, those containing less than a given percentage of divergent individuals (cut off = 2/n) and minimum allele frequency (MAF = 0.01), were removed using the informloci command in the poppr package (Kamvar et al. 2014).
Unique multi-locus genotypes (MLGs) were determined by first establishing the genetic distance between two sets of replicate samples to compensate for SNP genotyping errors associated with DArTseq. The average genotyping error for DArTseq SNPs between pairs of template control DNA samples has been reported as averaging 0.8%, which is lower than other SNP genotyping platforms and DNA sequencing (Ndjiondjop et al. 2018). Genetic distances within both sets of biological replicates were calculated using the provesti.dist model, and used to filter the data with the mlgfilter command in the poppr package (default parameters), such that isolates sharing a genetic distance below the calculated value were considered the same MLG. Isolates within the same MLG were classified as clones using the clonecorrect function in poppr.

Resolution of P. teres forms
Based on initial comparisons, DArTSeq SNP markers (rather than dominant silicoDArT markers) were chosen to assess the formae relationships, together with subsequent regional and field level analyses, due to the silicoDArT markers not adding novel information. SillicoDArT markers were analyzed separately, and generally supported the conclusions found with the DArT SNP markers (Supplementary File 1). An initial assessment of the effectiveness of DArT genotype data to distinguish P. teres formae isolates was conducted using a group of four Ptt isolates, including reference genome isolate W1-1, the Ptm reference isolate SG1 (Syme et al. 2018), and four randomly selected Ptm isolates (106/16A, 20P3001, PTM18-024, and 71/17, see Supplementary Table 2).

Genetic diversity and linkage disequilibrium
The genetic diversity of groups of isolates was determined by comparing the number of MLGs to the expected number of MLGs in the original data, together with Simpson's corrected index ((N/(N − 1))λ) of multi-locus genotype diversity. To determine the extent of random mating occurring within populations, gametic equilibrium was calculated using the standardized index of association (r d ), which is sample size independent (Agapow and Burt 2001), within the poppr package in Rstudio (RStudio, Boston, MA, USA).

Analysis of molecular variance
In order to detect genetic variation within and between groups of isolates for each comparison (between forms, regions, and fields), an analysis of molecular variance (AMOVA) was carried out. Variance (σ) as well as the population differentiation statistics (ϕ) were calculated using the poppr.amova command in the poppr package.

Principle components analysis and discriminant analysis of principle components
Principal component analysis (PCA), a multivariate statistical approach, was used to inform genetic clustering of individuals. In addition, genetic subdivision was inferred through discriminant analysis of principle components (DAPC), a method optimized for large datasets that like PCA does not rely on pre-existing genetic models but which can examine more complicated scenarios, detecting between-group variability and structures existing among clusters (Jombart et al. 2010). For the PCAs, clone-corrected data was investigated with the glpca function in the adegenet R package to observe the impact of eigenvalues on the overall variance explained (Jombart 2008). DAPC was performed after first observing the K-means clustering graph at the lowest Bayesian information criteria, using the find.clusters command in the adegenet package, and then applying the appropriate value for each of the DAPC analyses. DAPC was performed with the dapc command in adegenet, applying the result derived from the cross validation function Xval.dapc to confirm the correct number of PC to retain (Dray and Dufour 2007). The snpzip command using the "median" method in the adegenet package was then used to calculate the contribution (loadings) of each SNP to groups assigned in the clustering analysis. DArTSeq sequences (69 bp in length) responsible for genetic clusters were examined by BLAST at the National Center for Biotechnology Information and in Geneious 8.0.5 (Kearse et al. 2012) to find annotated homologous genes in Ptm or in related fungi including Ptt and Pyrenophora tritici-repentis (Ptr) (Supplementary Table 3).

Bayesian inference-based clustering analysis
Genetic structure was also examined via a Bayesian inferencebased method, implemented in STRUCTURE version 2.3.4 (Pritchard et al. 2000). STRUCTURE is a model-based clustering approach that assumes loci within populations are at Hardy-Weinberg equilibrium and linkage equilibrium, although allowing for admixture linkage disequilibrium within a population, enabling detection of subtle population subdivisions (Falush et al. 2003). SNP data was analyzed using a burn-in period of 100,000 steps and 100,000 replications, assessing K values between one and nine with 10 iterations. The STUCTURE output file data was then processed with STRUCTURE HARVESTER (Earl and vonHoldt 2011) to determine the optimal K value.

Genetic isolation by distance
To detect the correlation between genetic distance and geographic distance, a Mantel test was carried out based on each isolate's sampling GPS coordinates or, in the absence of these, the closest known location. The provesti.dist function was used to produce a matrix of the genetic distances between individuals and used alongside the dist function in R to produce a distance matrix of the GPS coordinates. The mantel.rtest was used to calculate the likelihood of isolation by distance in the ade4 package.

Inter-formae SNP diversity and differentiation
In order to assess the accuracy of the DArTseq SNP dataset to differentiate the P. teres forms and their relative genetic relatedness, subsamples of Ptm and Ptt individuals were compared. After data filtering, 1594 SNP markers remained. Total SNP variance within forma was relatively low (7.7%), comprising only a small portion of the total observed variation, whereas variation between formae was high (92.3%) and comprised the majority of the total variation. This was complimented by AMOVA differentiation statistics between the two groups, which was high (ϕ = 0.92), while a PCA showed the first two principle components explained ∼90% of the variation.

Field-level population structure of Ptm isolates
After filtering by allele call rate and missing data, 164 out of the initial 166 WA Ptm isolates remained, providing 11 to 36 isolates per field. Removing non-informative loci left 1252 SNP markers for further analysis. Total SNP variance within fields was high and comprised most of the total variation (≥99%), whereas variation between fields comprised ≤1% of the total. The isolates showed a high level of individual variation, grouping into 155 MLGs (Table 1), while population differentiation was low (ϕ = 0.0035). Nei's unbiased gene diversity, which indicates the probability that two randomly chosen alleles are different, was moderate for all populations (>0.196); and the corrected Simpson's index, which represents the probability of two random isolates drawn from a population to be of a different genotype, suggested high genotype diversity for all populations (1 − λd > 0.99). Despite the high genotype diversity, all fields appeared to be primarily reproducing asexually as suggested by the standardized index of association (r d ) values (between 0.003 and 0.019, P = 0.01).
A rarefaction curve showed no significant deviation in the number of MLGs found within each field based on non-clone-corrected data ( Supplementary Fig. 1), while a lack of curve saturation suggested a large number of MLGs exist at each sample site. Four MLGs contained between two and six individuals, with the remaining 151 MLGs consisting of a single representative. MLG 49 was the largest and most geographically widespread group, detected in three fields and comprising six individuals (Fig. 2).
Principal components analysis showed no observable population structure when the first two principal components were plotted against one another (Fig. 3). However, unsupervised clustering analysis performed without a priori knowledge of sample location suggested two potential populations (K = 2) when 150 principal components were retained. A membership probability plot of each of the fields illustrating the predicted populations suggested weak geographic ties to the predicted clusters ( Supplementary  Fig. 2), while all six fields contained isolates from cluster one and four contained isolates from cluster two.
Further analysis of these two putative populations by AMOVA revealed variation within clusters to be high (∼78%), however, variation between clusters was also moderately high (∼22%). A principle components analysis was performed on the putative  Data is presented for groups sampled within WA fields and across four regional areas, produced within the poppr package (Kamvar et al. 2014 Indicates the cumulative number of MLGs irrespective of field or region. two populations and showed significant differentiation (Fig. 4). The corrected Simpson's index also suggested high genetic diversity within each cluster (1 − λd > 0.98).
The role of Agzone origin or isolate host variety in genetic structure was also examined. Similar to the analysis at field level, variances within Agzones and within the same host cv were high and comprised the majority of the total variation (99%), whereas variation between Agzone and host cv was low (1%). MLGs containing multiple individuals were also shared between each Agzone and cv.

Inter-regional population structure analysis
Isolates collected for inter-regional comparisons were placed into four groups represented by between 8 and 30 individuals based on their origin of collection: Vic, SA, WA, with Qld and NSW combined. The isolates from NSW (n = 3) and Qld (n = 5) were combined as they represent a contiguous barley-growing area straddling the state border. After clone correction, SNP variance within regions was high and accounted for most of the variation (∼99.1%), whereas variation between the groups was low (∼0.9%). Regional differentiation of isolates was low, similar to that found at the field level (ϕ = ∼0.01). Isolates were highly heterogeneous at the regional level, presenting 85 MLGs based on 1278 loci with each MLG confined to a single region (Table 1). Nei's unbiased gene diversity was high for all regions (0.19-0.211), and the corrected Simpson's index suggested high genotype diversity (1 − λd > 0.99). Similar to comparisons between groups of isolates collected within fields, each of the four regional groups appeared to be primarily asexually reproducing as portrayed by the r̄d distribution.
PCA eigenvalues calculated for the first two PCAs linking samples to their sample locations accounted for a low ∼6% of the explained variation, with no visible population structure when plotted (Fig. 5). Unsupervised clustering analysis performed without a priori data of sample location suggested the existence of a unique population in Australia (K = 1). The Mantel test revealed a small, but positive linear correlation between genetic distance and isolate sample location (P = 0.024 and r = 0.073), suggesting that despite no detectable genetic structure being present between Australian regions, there is likely a component of isolation by distance (Supplementary Fig. 3).
To increase the data resolution supporting the putative populations established with DAPC in the field level analysis, isolates from both the field level and regional analysis were combined. For this, 239 clone-corrected Ptm isolates and 1,271 filtered SNP markers were used, in order to show the total MLG genetic diversity of the four sample regions (Table 2). Genotype diversity was high within each of four national Ptm groups and similar to the previous analyses, most of the variation (∼98.3%) occurred within regions, whereas variation between regions was low (∼1.6%), with variations between fields contributing least to genetic variation (∼0.1%).
Unsupervised clustering analysis by PCA performed without a priori knowledge of sample location suggested two potential populations (K = 2) ( Supplementary Fig. 4). DAPC was used to study the two putative populations with 32 PCA eigenvalues and three discriminant analysis eigenvalues being retained ( Supplementary Fig. 5). Notably, one of the populations (pop1) was present in all four regions; however, WA was the only region to contain the second population (pop2). Only one Western Australian isolate from the original regional analysis (19PTX147) contributed to the pop2, the other 11 were isolates from the field level analyses. Genetic diversity indices were high between each putative population in Table 2. Further analysis of the pop1 and pop2 groupings by AMOVA revealed variation within each to be high (∼79%), and the variation between moderately high (∼21%).
Validation of the two cryptic sub-groups with STRUCTURE version 2.3.4 (Pritchard et al. 2000) generally supported the DAPC findings, with both analyses showing no geographic ties of the isolates to their regions ( Supplementary Fig. 6). However, STRUCTURE indicated the presence of three potential populations rather than the two proposed by DAPC (Supplementary Fig. 7). This third subpopulation of just eight isolates (population assignment >50%) consisted of much more highly related individuals of the larger population identified in DAPC, which was also present only in WA ( Supplementary Fig. 8).

Mating type equilibria
To determine the level of sexual reproduction of Ptm in Australia, mating type analyses were performed. PCR amplification of MAT loci of the 87 interstate Ptm isolates revealed 42 that were MAT1-1 and 45 that were MAT1-2. Both mating types were present in each of the four state regions did not significantly deviate from a 1:1 ratio for a randomly crossing population in both the original and clone-corrected data (chi-square tests suggested no significant deviation from an expected 1:1 ratio at P > 0.05). Similarly, the isolates collected from individual WA fields did not significantly deviate from the 1:1 ratio, both within and between fields, in both the unaltered and clone-corrected data. The clonal isolates in the largest group of clones in MLG: 49, identified among WA isolates, congruently displayed the same mating type, MAT1-2 and, as may be expected, all clones within other MLGs amplified the same mating type.

SillicoDArT markers support lack of population structure and high genotypic diversity in Australia
Dominant SillicoDArT markers were examined independently to DArTSeq SNP markers in complimentary analyses of all regional and field Ptm isolates (Supplementary File 1). Results generally agreed with those based on the SNP markers. The SillicoDArT markers distinguished the same MLGs and supported the genetic diversity estimates, with most of the genetic variation (∼98%) occurring within regions, whereas variation between regions was low (∼1.4%), while variation between fields contributed least to genetic variation (∼0.6%). The silicoDArT markers also indicated high genotypic diversity but isolates were closely related.
Unsupervised clustering analysis primarily supported the SNP data, which established a group of isolates clustering separately from the other MLGs (Fig. 4: Cluster 2, Table 2: pop2). Notably, the same WA genotypic group was observed. Clustering analysis however differed in suggesting the existence of three other cryptic subpopulations unrelated to region, Agzone field, or host variety.

Population differentiating SNP markers are located in genomic regions associated with fungicide resistance
Following DAPC of all clone-corrected isolates, 18 unique markers were associated with differentiation of the two SNP-based genetic clusters (Supplementary Table 2). All the markers mapped by BLASTn to Ptm isolate SG1 (BioProject: PRJEB18107, BioSample: SAMEA4560037) within Geneious 8.0.5 (Kearse et al. 2012) with E values ranging from 3.92 × 10 −30 to 4.04 × 10 −10 with >95% identical sites shared. Of the 18 markers, 11 mapped to candidate genes and seven mapped to intergenic regions. The most differentiated marker was found at ∼8500 bp from the fungicide resistance-associated gene Cyp51A (GenBank accession: CP060577). Two other markers were also found within 16,000 bp and 31,000 bp of this gene. Two markers were found near to a Ptr homologue for a major facilitator superfamily (MFS) gene annotated as a benomyl/methotrexate resistance protein (GenBank accession: XM_001931024).

Discussion
This study combines the largest and most geographically diverse Australian collection of Ptm isolates to date with a high-resolution genetic marker technology, enabling an improved resolution of the genetic diversity and relatedness in the pathogen population. Overall, the study confirms high levels of genotype diversity and low levels of clonality. Similar results in both forms of P. teres have been observed previously in Australia (Bogacki et al. 2010;Ellwood et al. 2019) and in other countries such as Sweden (Jonsson et al. 2000), Italy (Rau et al. 2003), Czech Republic (Leisova et al. 2005), Finland (Serenius et al. 2007), Canada (Akhavan et al. 2016), and Iran (Vasighzadeh et al. 2021). Modest levels of genetic recombination were observed both within fields and within national regions. This was shown through linkage disequilibrium analysis using the standardized index of association, r d . For a freely recombining population, the expected score is zero and greater than zero if there is an association between alleles. Both categories had scores greater than zero, indicating limited sexual reproduction with asexual propagation the primary mode. However, equal mating type ratios and low numbers of clonal isolates were also observed, suggesting sexual reproduction is frequent enough to maintain mating type balance.

Lack of geographic population structure
Isolation by distance and physical barriers to migration commonly contribute to genetic differentiation between individuals. However, despite Australia being some 4000 km wide, geographic clustering of genotypes was absent in this study, with low variation between regions. Wind dispersal of spores provides the most likely explanation. However, the Nullarbor Plain is a mostly treeless, arid area, which stretches over 1000 km and separates WA from the rest of the barley-growing regions in Australia. Despite this, clustering analysis indicated a single Australian population (K = 1). Over larger distances, assisted dispersal in  Data is presented for combined regional Ptm collections and of two putative genotypic clusters defined by DAPC, produced within the poppr package (Kamvar et al. 2014). MLG: the total number of unique multi-locus genotypes (MLGs) identified per region or genotypic group. H: Shannon-Wiener index of MLG group genotypic diversity, a measure of the number of unique genotypes and their homogeneity. 1 − λ: corrected Simpson's index of MLG diversity, the probability that two isolates from the same dataset are different genotypes. H exp : Nei's unbiased gene diversity index, the probability that two randomly selected alleles are different. r d : the standardized index of association; with a value of zero for a null hypothesis, a population is freely recombining.
infected straw and hay may play a role in suppressing population differentiation, as these are commonly transported between states, particularly during droughts (Ptm is not regarded as a significant seed borne pathogen). Another contributing reason may be due to the diseases' relatively recent introduction into Australia, being first identified in Nabawa, WA in 1977 (Khan 1982), and a lack of time enabling divergence from a founder population. This is a less parsimonious explanation, and assumes similar genotypic lineages have established and persisted throughout the country. Nonetheless, despite the lack of obvious geographic population structure, a Mantel test suggested some isolation by distance, although the effect was low. Analysis of the population differentiation between fields in WA provided similar results to regional comparisons, with low genetic differentiation between individuals from different Agzones, fields, and the host variety, but a high degree within. This may be expected as, unlike comparisons of genotypes between different regions defined by clear geographic boundaries, there are extensive barley-growing regions in WA from Agzone 2 through to Agzone 6. Furthermore, the distribution of MLG groups indicated neither Agzone nor host cv appeared to play a role in genotypic selection among the isolates. The cvs the isolates were sampled from, RGT Planet and Spartacus CL, were independently developed by different breeding companies. Both are rated as susceptible to SFNB, and the Ptm genotypic compositions on these hosts do not indicate independent gain of virulence. These have been the most popular varieties sown in WA in recent years, whereby Spartacus CL was the most common variety in 2019 and 2020 with RGT Planet in second position, together accounting for nearly 70% of the area sown (Paynter et al. 2022). As other popular cvs are predominantly susceptible, the detection of gain of virulence by Ptm in Australia might be regarded as the exception. However, anecdotal reports suggest Ptm is capable of defeating SFNB resistance. For example in WA, Thomas et al. (2018) noted the erosion of partial seedling resistance in cvs Scope and Hindmarsh, while seedling resistance was overcome in cv Baudin (Muria-Gonzalez et al., manuscript submitted).
In the most intensively sampled region, WA, just three MLGs were found with individuals distributed across sites used for field level sampling at locations over 650 km apart. The low number of clones in these MLGs (11) provides weak evidence for recent selection of beneficial genotypes, perhaps masked by susceptible popular cvs. For example, predominantly clonal Ptm isolates were previously found in cv Oxford in southern WA, associated with a virulent new pathotype in combination with demethylase inhibitor (DMI) fungicide resistance (Turo et al. 2021). The transience of the genetic background of clonal expansions in Ptm may be similar to Ptt, where DMI resistance was rapidly assimilated into the wider population (Ellwood et al. 2019).

Genotypic clustering is independent of geographic origin
Clustering analysis by DAPC suggested the presence of two populations (K = 2). These were not linked to the field of origin, with individuals present across most of the sampled fields, potentially suggesting a recent introduction or selection of new genotypes variants in WA. Bayesian analysis with STRUCTURE suggested the presence of three populations (K = 3). The two populations identified by DAPC were supported, with a third composed of more highly related individuals representing a subset of the larger DPAC pop1 group. This population was present only in WA and may represent selection of an advantageous genotype, and contains MLG: 49, the largest group of clones found in WA.
However, despite the removal of isolates sharing the same MLG prior to STRUCTURE analysis, the validity of this group should be assessed in the context of the limitations of the STRUCTURE algorithm, which requires populations to be in Hardy-Weinberg equilibrium.
A lack of Australian Ptm geographic population structure is in agreement with a study by McLean et al. (2014), which assessed the genetic structure of Ptm in Australia using a geographically diverse set of isolates collected between 1996 and 2009 (n = 60). The authors also identified two genetically distinct Ptm clusters unlinked to sampling year and sampling region, however, in that study an Australia-wide population structure was not established. Our study confirms the presence of a single Australia-wide Ptm population, with one or more smaller populations unrelated to sampling site, referred to here as cryptic populations.
Our results also complement aspects of other Ptm genetic diversity studies conducted within Australia. Bogacki et al. (2010) found that isolates sampled from different areas of the same field and between fields were genetically similar and Serenius et al. (2007) found that the majority of the genetic differentiation of Ptm occurred within fields rather than between fields or regions. However, these two studies differ in their degree of differentiation, Bogacki et al (2010) finding a very low degree of differentiation between Ptm field populations in line with the results of this study whereas Serenius et al. (2007) finding a much larger degree between states and fields (21.94 and 11.24%, respectively) with low variation within fields (66.82%). In this study, 98% if the genetic variation occurred within fields and only 1.6% occurred between fields and/or regions.
There are different explanations for the contrasting results between our study and previous research; greater distances between sampling sites should mean less migration, and therefore larger differences should be found between sampling locations than within the same fields. However, when comparing fields from Agzone 2 and Agzone 6, which are 650 km apart from each other, genetic diversity within fields is significantly higher than between fields. A more likely explanation for the differing results is the molecular marker technologies used to differentiate isolates, with the major factor affecting resolution appearing to be a comparatively low number of genetic markers in the earlier studies.

Genotypic clustering is associated with fungicide resistance
Three of the markers identified by DAPC associated with population differentiation were located on Chromosome 6, within 31 Kbp of the DMI fungicide resistance-associated gene Cyp51A (Mair et al. 2020) and two of these markers contributed most to differentiation. A second group of markers implicated with fungicide resistance were located within 3 Kb of a Ptr orthologue encoding a benomyl/methotrexate resistance-like protein, and may indicate an alternate mode of fungicide resistance in Ptm. Related genes are implicated in multi-drug resistance (MDR) and are part of the MFS. MDR transporters are proton antiporters that mediate the efflux of a diverse range of drugs and toxic compounds. For example, they have been implicated in providing resistance to quinidine (Vargas et al. 2007), amiloride (Stolz et al. 2005), and fluconazole (Keniya et al. 2015) in yeast. In the wheat pathogen, Zymoseptoria tritici, MDRs provide enhanced fungicide resistance to tolnaftate, terbinafine, the DMI metconazole, the quinone outside inhibitor azoxystrobin, and the succinate dehydrogenase inhibitor boscalid (Omrane et al. 2017). These marker associations concur with an international study of Ptt, where markers nearby to Ptt Cyp51A and a MFS domain-containing protein were also implicated in underlying population structure (Dahanayaka et al. 2021).
Other markers that were associated with genetic structuring included a predicted Ptm non-ribosomal peptide synthase (NRPS) gene that showed homology to a Ptr gene annotated as HC-toxin. HC-toxins are a form of host-selective toxins that are thought to operate through the prevention of defense gene expression. For example, HC-toxin is a tetrapepetide produced by Cochliobolus carbonum, which inhibits histone deacetylase leading to disease symptoms on susceptible plants (Brosch et al. 1995). However, the predicted product of the Ptm gene in question is a likely a pentapeptide based on the number of NRPS gene modules, therefore the function may be different. Other cyclic pentapeptide host-selective toxins such as victorin, produced by Cochliobolus victoriae, are known to exist, however these are not currently considered NRPSs (Kessler et al. 2020). NRPSs also produce other bio-active secondary metabolites, such as those involved in cellular development and stress response (Keller et al. 2005). The possible role of this gene in differential virulence of Ptm isolates may warrant further examination in future studies.
Additional Ptm genes orthologous to annotated Ptr genes included: A calcium channel protein 1 (CCH1), which may have a role in homeostasis and virulence in Aspergillus fumigatus (de Castro et al. 2014); tRNA A64-2′-O-ribosylphosphate transferase protein, which modifies cytoplasmic initiator tRNAs, preventing them from participating in translational elongation (Kiesewetter et al. 1990); finally, a pumilio domain-containing protein, implicated in post-transcriptional regulation influencing mRNA stability, translation, and localization (Wang et al. 2018).
This study has revealed a single Australia-wide population of Ptm suggesting unconstrained movement of genotypes within Australia. There also appears to be a lack of recent directional selection although we detected cryptic populations. This is consistent with the most popular barley varieties being susceptible to SFNB (Paynter et al. 2022), while the cryptic populations appear to be related to generalized adaptions, in particular the widespread application of fungicides, but may include other environmental or biotic stresses.
Incorporating new resistance genes into modern cvs is the most cost-effective approach to combatting serious crop diseases such as SFNB. This research is of significance to breeders and growers alike, as long-term control measures require an improved understanding of plant-pathogen interactions. Ptm is genotypically diverse and mobile pathogen, underlying a need for nationally coordinated disease management and resistance breeding strategies. The challenge to breeders in developing new varieties will be in extensive testing against the range of known pathotypes and in deploying resistance genes combinations that reduce the likelihood of their breakdown.

Data availability
Supplemental Material included at figshare: https://doi.org/10. 25387/g3.21611043. Supplementary Table 1 contains DArTseq SNP and dominant markers together with metadata descriptions and is hosted at: https://figshare.com/s/c2dcabcf2351a9043e30. Supplementary Table 2 gives isolate collection site data, forma and mating type of P. teres isolates. Supplementary Table 3 gives Ptm DAPC SNP markers contributing to population structure and annotations of corresponding orthologous genes in Ptr.
SilicoDArT analyses are provided in File S1. Supplementary Figures 1-8 are combined in a single file with the following contents: Supplementary Fig. 1. Rarefaction curve analysis describing observed Ptm MLG richness. Supplementary  Fig. 2. Membership probability plot showing the population assignment for each of the 155 field isolate MLGs and their respective collection sites. Supplementary Fig. 3. Isolation by distance plot illustrating the pattern of genetic differentiation among Ptm genotypes collected from four Australian regions. Supplementary Fig. 4. Discriminant analysis of principal components (DAPC) cluster analysis of Australian Ptm isolates. Supplementary Fig. 5. Scatter plot of the first two PCA principle components resolving Australian Ptm isolates. Supplementary  Fig. 6. Bayesian cluster analysis of Australian Ptm genotypes. Supplementary Fig. 7. Structure Harvester cluster analysis of Australian Ptm isolates. Supplementary Fig. 8. Scatter plot of the first two principle components based on data from STRUCTURE 2.3.4.

Funding
This research was funded under Grains Research and Development Corporation (GRDC) PhD scholarshipCUR2002-002RSX.

Conflicts of interest statement
The author(s) declare no conflict of interest.