Global diversity and distribution of three necrotrophic effectors in Phaeosphaeria nodorum and related species


Author for correspondence:

Megan C. McDonald

Tel: +61 261255561



  • Population genetic and phylogenetic studies have shown that Phaeosphaeria nodorum is a member of a species complex that probably shares its center of origin with wheat (Triticum aestivum and Triticum durum). We examined the evolutionary histories of three known necrotrophic effectors (NEs) produced by P. nodorum and compared them with neutral loci.
  • We screened over 1000 individuals for the presence/absence of each effector and assigned each individual to a multi-effector genotype. Diversity at each NE locus was assessed by sequencing c. 200 individuals for each locus.
  • We found significant differences in effector frequency among populations. We propose that these differences reflect the presence/absence of the corresponding susceptibility gene in wheat cultivars. The population harboring the highest sequence diversity was different for each effector locus and never coincided with populations harboring the highest diversity at neutral loci. Coalescent and phylogenetic analyses showed a discontinuous presence of all three NEs among nine closely related Phaeosphaeria species. Only two of the nine species were found to harbor NEs.
  • We present evidence that the three described NEs of P. nodorum were transmitted to its sister species, Phaeosphaeria avenaria tritici 1, via interspecific hybridization.


Necrotrophic fungal pathogens secrete a variety of necrotrophic effectors (NEs) (syn: host-specific/selective toxins) that interact in a gene-for-gene manner with host susceptibility genes (Oliver & Solomon, 2010). Typical plant defense responses involve the induction of pathogenicity-related genes, which leads to the production of antimicrobial compounds, the accumulation of reactive oxygen species and localized cell death. This defense response is known as the hypersensitive response (HR) ( Jones & Dangl, 2006). A growing body of evidence supports the hypothesis that necrotrophic pathogens have taken advantage of the HR, using small secreted proteins or secondary metabolites to activate HR preceding fungal growth (Effertz et al., 2002; Liu et al., 2009, 2012; Lorang et al., 2012). This group of molecules is collectively referred to as NEs.

Effectors are a class of pathogen proteins or metabolites whose function is to alter or suppress the host's normal immune response. Three NEs have been described for the fungal wheat pathogen Phaeosphaeria nodorum (Friesen et al., 2006; Liu et al., 2009, 2012). Each of these NEs is a small, secreted protein that displays a presence/absence polymorphism in natural field populations. SnTox1 is a cysteine-rich protein that was shown to exhibit significant diversifying selection (Liu et al., 2012). SnTox3 has no known homology to any proteins available in public databases (Liu et al., 2009). SnToxA has limited homology to a prokaryotic gene and also exhibits significant diversifying selection (Stukenbrock & McDonald, 2007). Transformation with any of the three NEs into a nonpathogenic fungal isolate was sufficient to induce necrosis on susceptible wheat (Triticum aestivum) cultivars (Friesen et al., 2006; Liu et al., 2009, 2012). Clamped homogeneous electric field (CHEF) gel analysis indicates that each NE is located on a different chromosome and each gene is located on a different scaffold in the genomic assembly (Hane et al., 2007; Liu et al., 2012).

While there is a growing list of shared properties associated with effectors (reviewed in Kamoun, 2007 and Stergiopoulos & de Wit, 2009), very little is known about the evolutionary origins of effector-encoding genes. Recent genome sequencing has revealed families of what have been termed core effector proteins within the Dothideomycetes (Stergiopoulos et al., 2010, 2012). Identifying these core effectors relies on conservation of homologous elements within the effector proteins. The three characterized NEs of P. nodorum share no homology with any proteins for any fungal species available in GenBank and they do not appear to represent core effector proteins. Population genetic studies of effector loci have provided important insights into the evolutionary processes that affect NE loci within a species. A population genetic analysis of the NIP1 gene in Rhynchosporium commune showed that alteration of the effector protein sequence or deletion of the effector allele could lead to virulence (Schürch et al., 2004). Studies on the obligate biotroph of flax (Linum usitatissimum), Melampsora lini, revealed high levels of nonsynonymous substitutions at the AvrL567 locus and sectional insertions/deletions at the AvrM locus that alter or abolish recognition by their corresponding resistance (R) genes (Dodds et al., 2006; Ellis et al., 2007). Virulence alleles in Leptosphaeria maculans were attributed to both deletion of the AvrLm6 locus and introduction of early stop codons by repeat-induced point mutation (RIP) (Fudal et al., 2009; Van de Wouw et al., 2010).

For NEs, deletion of the effector gene leads to loss of the virulent phenotype on hosts with compatible genetic backgrounds. To date, effector studies have focused on individual genes or small groups of effector genes in a small number of individuals (Dodds et al., 2006; Barrett et al., 2009; Liu et al., 2009; Chuma et al., 2011; Liu et al., 2012), though it is clear that fungal populations are large and capable of harboring high levels of NE diversity. As more effector genes are discovered and characterized, a key question has become: how did this class of genes originate within fungal pathogens? Understanding the evolutionary origins of these genes could provide significant insights into the mechanisms involved in pathogen emergence and host specificity.

The horizontal transfer of SnToxA from P. nodorum to Pyrenophora tritici-repentis is thought to have led to the emergence of P. tritici-repentis as the tan spot pathogen on wheat (Friesen et al., 2006). Detection of this horizontal gene transfer (HGT) event was made possible by the high sequence similarity between the SnToxA alleles found in P. nodorum and the PtrToxA allele found in P. tritici-repentis. It remains unknown if the NEs present in P. nodorum are the result of a long co-evolutionary process between the pathogen and its hosts, or alternatively if the NEs were acquired more recently via other mechanisms such as horizontal transfer or interspecific hybridization.

This study focuses on the population genetics and evolutionary history of SnTox1, SnTox3 and SnToxA in P. nodorum and its closest known relatives. We assessed the global distribution and geographic diversity for all three effectors. We determined the presence or absence of each gene in over 1000 global isolates using both PCR and Southern hybridization. We calculated the frequency of each NE over spatial scales ranging from fields to continents and generated multi-effector genotypes to determine if selection was operating on the combination of NEs. To gain insight into the ancestral origin of these NEs in P. nodorum, we assessed the presence or absence of each NE in eight recently described sister species. Finally, we sequenced each NE in several hundred global strains to compare NE sequence diversity with previously published population genetic studies based on neutral markers.

Materials and Methods

Isolate collection

Isolates used in this study are described in Table 1. Fungal hyphae were transferred to 50 ml of yeast sucrose broth (YSB; 10 g l−1 yeast extract and 10 g l−1 sucrose) and grown on a rotary shaker for 3 d at 120 rpm at 18°C. Fungal biomass was lyophilized and ground into a powder and total DNA was extracted using the DNeasy Plant Mini DNA extraction kit (Qiagen GmbH, Hilden, Germany) following the manufacturer's instructions. Extracted DNA was diluted 1 : 500 with sterile water for PCR amplification.

Table 1. List of regional collections, sample sizes and host source for all isolates by species
RegionsYearCollectorLocation n Host source
Phaeosphaeria nodorum
Iran2005R. SommerhalderGolestan Province27Wheat (ears)
2010M. RazaviGolestan Province13Wheat (ears)
Central Asia2005R. SommerhalderAzerbaijan7Wheat
2003–2004H. Maraite, E. DuveillerKazakhstan32Wheat
2004H. Maraite, E. DuveillerTajikistan7Wheat
2003H. Maraite, E. DuveillerKyrgyzstan1Wheat
2003H. Maraite, E. DuveillerRussia9Wheat/durum wheat
2005M. AbangSyria2Wheat
China2001R. WuFujian Province115Wheat
Europe1992M. ShawEngland9Wheat
2005E. StukenbrockDenmark54Triticale/wheat
2005E. BlixtSweden55Wheat
1994S. KellerSwitzerland113Wheat
USA1991G. Milus, M. GrayArkansas94Wheat (seed)
1993G. ShanerIndiana8Wheat
1993L. FranclNorth Dakota12Wheat
1979J. KrupinskyNorth Dakota2Agropyron spp.
1998J. KrupinskyNorth Dakota9Barley
2003P. LipsOH03 Sn15011Wheat
2006T. FriesenNorth Dakota6Wheat
1992G. BergstromNew York46Wheat
1993M. SchmidtOregon92Wheat
1992B. McDonaldTexas92Wheat
South Africa1994P. CrousSouthwestern Cape74Wheat
2007Z. PretoriusWestern Cape112Wheat
Australia2001B. McDonaldNarrogin73Wheat
Total P. nodorum   1065 
Phaeosphaeria avenaria tritici 1
Canada1991R. ClearAlberta65Wheat (seed)
1991R. ClearManitoba7Wheat (seed)
1991R. ClearSaskatchewan36Wheat (seed)
USA1993L. FranclNorth Dakota12Wheat
2005T. FriesenNorth Dakota8Wheat
Iran2005M. RazaviGolestan Province6Wheat (ears)
2010M. RazaviGolestan Province18Wheat (ears)
Total Pat1   152 

Data collection

PCR amplification was performed in 20-μl reactions containing 0.05 μM of each primer (Microsynth, Balgach, Switzerland), 1× Dream Taq Buffer (MBI Fermentas, Amherst, NY, USA), 0.4 μM dNTPs (MBI Fermentas) and 0.5 units of Dream Taq DNA polymerase (MBI Fermentas). The PCR cycle parameters were: 2 min of initial denaturation at 96°C followed by 35 cycles of 96°C for 30 s, annealing for 45 s and extension at 72°C for 1 min. A final 7-min extension was performed at 72°C. PCR products were purified to remove unincorporated nucleotides and primers using NucleoFast® 96 PCR plates (Macherey-Nagel, Oensingen, Switzerland). Details of the annealing temperature and primers used have been published previously; see Friesen et al. (2006) for SnToxA, Liu et al. (2009) for SnTox3 and Liu et al. (2012) for SnTox1.

Sequencing reactions were conducted in a 10-μl volume using the BigDye® Terminator v3.1 Sequencing Standard Kit (Life Technologies, Applied Biosystems, Grand Island, NY, USA) with both the forward and reverse primers. The cycling parameters were 96°C for 2 min followed by 55 or 99 cycles of 96°C for 10 s, 50°C for 5 s and 60°C for 4 min. PCR products were cleaned with the illustra™ Sephadex™ G-50 fine DNA Grade column (GE Healthcare, Pittsburgh, PA, USA) according to the manufacturer's recommendations and sequenced with a 3730xl Genetic Analyzer (Life Technologies, Applied Biosystems). Alignment of forward and reverse sequences was performed in SeqScape software V2.5 (Life Technologies, Applied Biosystems). Quality screening and ambiguous base calls were edited by hand using SeqScape. Final alignments were exported and re-aligned using ClustalW, implemented online using the Max Planck Institute of Bioinformatics Toolkit ( Sequences that have been published previously are SnToxA accession numbers DQ423483; EF108451EF108463; SnTox3 accession number FJ823644; SnTox1 accession numbers JN791682JN791693. New sequence haplotypes described in this manuscript are deposited in GenBank under accession numbers JX997397JX997421.

Genomic DNA was blotted onto nylon membranes using a Bio-Dot microfiltration apparatus (Bio-Rad) following the instructions in the user manual. Radioactive probes were synthesized using purified PCR product (purification using NucleoFast® 96 PCR plates). Probe labeling with 32P was performed with 25–50 ng μl−1 of template DNA using the NEBlot Kit (NE Biolabs, Ipswich, MA, USA) following the manufacturer's recommendations. Unincorporated nucleotides were filtered out using illustra NICK™ columns (GE Healthcare). DNA hybridization, membrane washing and image acquisition were performed as described previously (McDonald & Martinez, 1990).

Data analysis

Contingency χ2 tests and Fisher's exact tests

Effector presence or absence was summarized as allele frequencies within each population. Contingency χ2 tests were used to test for independence of effector frequencies between populations. The program chifish was used to calculate both Fisher's exact test and Pearson's χ2 test for each locus separately. P-values for independence based on the combined effector frequencies were calculated by summing the χ2 test statistics and degrees of freedom, or by combining the P-values from Fisher's exact test using Fisher's method (Ryman & Jorde, 2001; Ryman, 2006). A Bonferroni correction was applied to correct P-values for multiple testing. After confirming the independence of each population, the expected number of multilocus genotypes was calculated. The frequency of each effector within a population was used as the expected probability of sampling an individual carrying the corresponding allele. For three bi-allelic loci there are eight possible multilocus genotypes (ToxA+Tox3+Tox1+; ToxA–Tox3–Tox1– etc.). The expected number of multilocus genotypes was compared with the observed number of multilocus genotypes using Fisher's exact test in each population (Fisher, 1925). This test was deemed to be most appropriate for large multinomial data sets with small expected values (n < 5) (Maiste & Weir, 2004). Fisher's exact tests were performed using the statistics package implemented in R. For contingency tables larger than 2 × 2, an estimation of the P-value was made using Monte Carlo simulations. Our test was conducted with MonteCarlo = TRUE with 10 000 replicates (Patefield, 1981). Each population was tested for independence separately, so no P-value corrections were applied.

Gene diversity and haplotype networks

Effector sequences were collapsed into haplotypes using Map, implemented in the Java-based program package snap workbench (Price & Carbone, 2005; Aylor et al., 2006). This program implements several individual programs into one platform in order to facilitate analysis of population parameters (Price & Carbone, 2005). The haplotype alignment was used in tcs 1.3 to generate the most parsimonious haplotype network (Clement et al., 2000). This program utilizes statistical parsimony methods to infer unrooted cladograms based on Templeton's 95% parsimony connection limit. Mutational steps resulting in nonsynonymous changes were identified manually using Sequencher v4.8 (Gene Codes Corp., Ann Arbor, MI, USA).


Rarefaction was used to identify regions of highest sequence diversity among populations with unequal sampling sizes. Rarefaction analysis was conducted using the method described by Simberloff (1978) and implemented in an online calculator ( Rarified sample sizes from 10 to 20 were used to estimate the number of sequence alleles by global region. The output given is the rarefied allele count and standard deviation at each of the given sample sizes. Standard deviations were used to calculate the standard error (SE). The number of sequence alleles and the 95% confidence interval (1.96 × SE) were plotted for each sample size and population using R.


Frequencies of NEs differ among populations but are randomly assorted within populations

Effector presence or absence was determined with gene-specific PCR primers for over 1000 isolates across seven major regions, including North America, Europe, Iran, Central Asia, China, South Africa and Australia (Table 1). Results from PCR assays were compared to Southern hybridization assays using 193 isolates for SnToxA, 284 isolates for SnTox3 and 242 isolates for SnTox1. The average disagreement between the PCR and hybridization assays across all three loci was 4%. Sixteen out of the 30 discrepancies involved the highly polymorphic SnTox1 locus and six of these were in the Central Asian population. When the two assays were not in agreement, the result from the hybridization assay was included in further analyses. Effector frequencies were calculated for each region and for the 16 subpopulations; variances were estimated using 100 bootstrapped replicates with replacement by region (Fig. 1). SnToxA was present at a lower global frequency than the other two effectors. SnTox1 was found at a frequency > 60% in all continental regions. All three effectors were present at high frequencies in South Africa and Australia.

Figure 1.

Frequency of Phaeosphaeria nodorum necrotrophic effectors (NEs) is different between regions. Bar plots show the frequency of each NE within P. nodorum in seven global regions. Bootstrap re-sampling with replacement was performed on each data set to generate variance for the mean frequency of each NE. 95% confidence intervals are plotted for the bootstrap replicates.

Effector frequencies differed significantly among field populations for all three loci according to contingency χ2 tests and Fisher's exact tests (Supporting Information Table S1). In many cases, effector frequencies among field populations within a region (e.g. Arkansas and Texas within North America or Sweden and Denmark within Europe) also differed significantly.

The complete multilocus effector presence/absence genotype was determined for 945 strains. Given three effector loci, there are eight possible multilocus presence/absence genotypes. Based on the observed effector frequencies in each population, the expected number for each multilocus genotype under random mating was estimated and compared to the observed number of multilocus genotypes (Fig. 2). Fisher's exact test was used to determine if the number of observed genotypes differed from the number of expected genotypes. The only population with a significant deviation from expected proportions was Oregon (Table 2). Values for European and North American populations were summed to generate the graph shown in Fig. 2. Sample sizes for each NE locus are given along with the number of isolates for which complete genotypes were scored (Fig. 2).

Table 2. Fisher's exact tests for observed versus expected number of genotypes by population
  1. a

    Number of observed genotypes deviates significantly from the number of expected genotypes.

Iran 1.00
Central Asia 1.00
China 1.00
North Dakota1.00
New York0.27
South Africa 0.13
Australia 0.43
Figure 2.

The observed number of multi-effector genotypes matches the expected number under an assumption of random mating in Phaeosphaeria nodorum. The total number of P. nodorum isolates scored for the presence/absence of each locus is shown under N and alongside is the per cent presence of each locus in the sample. The bar plots show the observed multi-effector genotypes compared to the expected number of genotypes calculated from the effector presence/absence data. The total number of individuals for which all three necrotrophic effectors (NEs) were scored is listed as n = in the top right corner of the bar plot.

Effector sequence alleles are diverse and globally distributed

We sequenced SnTox1, SnTox3 and SnToxA from 295, 203 and 178 individuals, respectively. The 295 SnTox1 sequences collapsed into 18 nucleotide haplotypes, which encoded 14 different proteins (Fig. 3a). The SnTox1 haplotype network contains three loops, indicating that some of the alleles originated through intragenic recombination. SnTox3 collapsed into 11 nucleotide haplotypes encoding six different proteins (Fig. 3b). The 178 sequences from SnToxA collapsed into 17 unique haplotypes. Two SnToxA alleles with nonsense mutations were detected, one reported earlier in South Africa (Stukenbrock & McDonald, 2007) and a second one found in New York. Excluding these two nonsense haplotypes, there were nine different SnToxA proteins (Fig. 3c). Evidence for intragenic recombination was also found at the SnToxA locus.

Figure 3.

Haplotype networks for each necrotrophic effector (NE) and sample sizes for sequence data by region. (a) The SnTox1 haplotype network for the 18 sequence alleles found in Phaeosphaeria nodorum and Phaeosphaeria avenaria tritici 1 (Pat1). Pat1 haplotypes are drawn in red; all other haplotypes are P. nodorum. Circle sizes reflect the frequency of each haplotype and colors correspond to the global region where it was found. Haplotype numbers, denoted ‘H#’, correspond to previously published sequences. Newly described haplotypes are marked with §. Nonsynonymous mutations are marked with asterisks and synonymous mutations or mutations in the intron are labeled with ‘S’ and ‘I’, respectively. Open circles represent missing haplotypes in the network. Loops in the network indicate potential intra-locus recombination events. Unique features of the network are labeled with red lines or boxes and a short description. Black brackets indicate mutations that occurred within the same codon. (b) The SnTox3 haplotype network. (c) The SnToxA haplotype network. The two nonsense haplotypes are labeled with a red hexagon.

Effector sequence diversity does not correspond with the center of diversity for neutral markers in P. nodorum

The region with the highest diversity for neutral markers in P. nodorum corresponds with the acknowledged origin of its wheat host in the ancient Fertile Crescent (Stukenbrock et al., 2006; Balter, 2007; McDonald et al., 2012). Allele diversity for each effector locus was compared among populations using rarefaction analysis across sample sizes ranging from 10 to 20 individuals. The rarefied number of alleles and 95% confidence intervals for each sample size are shown in Fig. 4(a–c). For SnToxA the South African population clearly contained the highest number of alleles. For SnTox1 both Europe and North America had the highest number of alleles. The average number of alleles found within each population was much lower for SnTox3, with North America and Australia showing the highest number of alleles. While the highest diversity based on neutral sequence loci and microsatellite loci was found in Iran (McDonald et al., 2012), the highest diversity for each effector locus was always outside of Iran (Fig. 4). The number of private alleles (alleles found in only one region) is summarized in Fig. 4(d). Private sequence alleles comprised 56% (10 out of 18) of the SnTox1 haplotypes, 64% (7 out of 11) of SnTox3 haplotypes and 76% (13 out of 17) of SnToxA haplotypes.

Figure 4.

The rarefied sequence allele count for each necrotrophic effector (NE) and number of private sequence alleles within each global region. (a–c) The mean number of alleles, after rarefaction, for sample sizes ranging from 10 to 20 randomly chosen individuals from each population. Vertical bars represent the 95% confidence interval for the mean number of alleles. All lines are Phaeosphaeria nodorum isolates, except for red lines with open diamonds, which are Phaeosphaeria avenaria tritici 1. (d) Pie charts showing the number and geographic location of private sequence alleles by NE locus.

Phylogenetic distribution of NEs among Phaeosphaeria spp

Phaeosphaeria nodorum is one of nine members of a species complex infecting wheat and other grasses (McDonald et al., 2012). The presence or absence of the three NEs was determined in all nine species using both PCR and Southern hybridization. The three effectors were found only in P. nodorum and Phaeosphaeria avenaria f. sp. tritici 1 (Pat1). Fig. 5 shows the coalescent multilocus species trees (adapted from McDonald et al., 2012) for the nine Phaeosphaeria species as well as the frequency of each effector within P. nodorum and Pat1. Only one Pat1 isolate carried SnTox3 (Fig. 5).

Figure 5.

Two out of nine Phaeosphaeria species carry necrotrophic effectors (NEs). The multilocus maximum-clade credibility tree adapted from McDonald et al. (2012) is shown. The tree is a coalescent tree with relative time on the x-axis and posterior probabilities for the branches shown. The two species within which NEs were found were Phaeosphaeria nodorum and Pat1. Other species are Phaeosphaeria avenaria tritici 3, 4, 5 and 6, Phaeosphaeria avenaria and unnamed species Phaeosphaeria 1 and 2, described in McDonald et al. (2012). The number of NE sequence alleles (n alleles) is shown above. The per cent NE presence within the species is shown below.

The number of sequence alleles for each effector in both species is shown in Fig. 5. The 37 SnTox1 sequences from Pat1 collapsed into a single haplotype that was shared with both North American and Iranian P. nodorum isolates. The single SnTox3 sequence found in Pat1 was also the most frequent P. nodorum haplotype. Three SnToxA alleles were found among the 14 Pat1 isolates sequenced. One SnToxA haplotype sequence was unique to Pat1 but had the same amino acid sequence as the most frequent effector allele in the SnToxA network. The other two Pat1 alleles were found in the two most common P. nodorum haplotypes. The overall diversity of Pat1 sequence alleles was far lower than found in P. nodorum, with all but one of the alleles identical in sequence to a common P. nodorum allele. This pattern is consistent with the hypothesis that Pat1 acquired all three NEs recently from P. nodorum.


Phylogenetic analysis of the Phaeosphaeria species complex revealed that only two out of nine closely related species carry the NEs SnTox1, SnTox3 and SnToxA. All three NEs are diverse at the amino acid level with a high proportion of population-specific sequence alleles. Rarefaction analysis indicated that the center of diversity for each NE did not correspond with previous population genetic studies that identified the highest levels of diversity at neutral genetic loci in Iran (McDonald et al., 2012).

Our analyses showed that the likelihood of an individual carrying one of the three NEs is population dependent. Phaeosphaeria nodorum is a sexual pathogen with a large effective population size that exhibits high levels of gene flow over continental scales (e.g. between Oregon, Texas and New York within North America; Keller et al., 1997; Stukenbrock et al., 2006). Based on gene flow estimates from neutral microsatellite loci, the expected frequencies of NEs in pathogen populations are not expected to differ among populations within a continent. Instead, we found significant differences in NE frequencies among many field populations that did not differ for neutral markers (Table S1). We believe that the differences in NE frequency among populations reflect differences in the frequencies of the corresponding host NE sensitivity genes among regions. As already shown for SnToxA and the corresponding host sensitivity protein Tsn1, we hypothesize that the activity of SnTox1 and SnTox3 depends upon an interaction with a corresponding host sensitivity protein which is present in some wheat cultivars but absent in others (Friesen et al., 2006; Liu et al., 2009, 2012). Unless there is a secondary virulence function, in the absence of a host sensitivity protein, pathogen strains carrying the corresponding NE have no fitness advantage and the NE gene is expected to be essentially neutral and subject to genetic drift (Tan et al., 2012).

This highlights one of the main challenges associated with controlling globally disseminated pathogens with a high capacity for gene flow. Susceptible cultivars planted within the dissemination range of pathogen populations carrying an NE could rapidly select for pathogen populations carrying the NE. Thus, breeding efforts should be coordinated across large geographic regions to eliminate known susceptibility genes and reduce the frequencies of NEs at the continental scale. This type of effort is now underway in Australia to eliminate the Tsn1 gene that encodes susceptibility to SnToxA and the sensitivity locus for Tox3 (Oliver & Solomon, 2010; Waters et al., 2011).

A similar study that illustrated the dynamics of multiple effector loci in large natural populations was recently completed using the flax rust pathogen Melampsora lini. Thrall et al. (2012) found dramatic fluctuations in the frequency of M. lini avirulence alleles across multiple loci and they were able to correlate these fluctuations with the susceptibility of the corresponding host populations. Their analyses show how rapidly the genotype frequencies of host and pathogen can change in a gene-for-gene system experiencing antagonistic co-evolution. While we did not measure the sensitivity of the host in each field population, the large differences in local NE frequency despite high levels of neutral gene flow suggest that there is very strong selection operating on NEs at the field level.

Despite significant differences in NE frequency between field populations, the distribution of multi-effector genotypes within all but one of the 16 field populations did not differ from the expectation under random mating of neutral markers (Table 2, Fig. 2). Traditionally, pathogens carrying particular combinations of avirulence or effector genes are classified into races (Barrett et al., 2009), analogous to the multi-effector genotyping conducted in this study. The ‘cost of carrying’ is believed to drive the loss or alteration of the avirulence gene. Our finding of random associations among effector alleles within a population suggests that there is little fitness cost associated with carrying particular combinations of NEs. We hypothesize that the ‘carrying cost’ of these effector genes is low in the absence of the corresponding host sensitivity allele. This finding is also consistent with the hypothesis that host cultivar is the main determinant of NE frequency in these pathogen populations.

The haplotype networks presented in Fig. 3 show a prevalence of nonsynonymous mutations. It was reported previously that SnToxA and SnTox1 are under significant positive selection (Friesen et al., 2006; Liu et al., 2012), while SnTox3 did not show evidence of positive selection (Liu et al., 2009). The detection of positive selection operating on protein-coding genes with unknown function has become a powerful tool for identifying effectors in fungal genomes (Ma & Guttman, 2008; Raffaele et al., 2010; Stukenbrock et al., 2011; Saunders et al., 2012). It is often assumed that the higher rates of nonsynonymous substitutions exhibited by pathogen effector genes reflect diversifying selection favoring novel effector variants that are not recognized by plant R proteins. Dodds et al. (2006) performed experiments that supported this assumption in the gene-for-gene interactions between M. lini and flax, where specific amino acid changes in Avr proteins altered or abolished recognition by their corresponding R proteins. For necrotrophic pathogens, it was hypothesized that the inverse process was operating, whereby higher rates of nonsynonymous substitution within NE proteins resulted from the pathogen tracking changes in the host susceptibility alleles (Stukenbrock & McDonald, 2007). An alternative hypothesis to explain the higher rates of nonsynonymous substitution seen in NEs is that positive selection has favored mutant NE alleles that increase pathogen fitness through a quantitative increase in virulence.

This alternative hypothesis is supported by recent experimental studies showing that the most frequent SnToxA protein variant is significantly more active against identical wheat Tsn1 alleles hypothesized to increase pathogen fitness (Tan et al., 2012). The haplotype networks of both SnTox1 and SnTox3 exhibit two or more frequent and widely distributed protein variants that differ at two or more amino acid positions. We hypothesize that the most common protein variants in these networks induce significantly more necrosis than the less common protein variants in the network. Experimental testing of this hypothesis is now underway.

There has been a rapid expansion of literature describing fungal effectors as small, secreted proteins or small metabolites that interact with the host to suppress or alter the immune response (reviewed by Kamoun, 2007; Hogenhout et al., 2009; Stukenbrock & McDonald, 2009; de Wit et al., 2009). As more effectors are identified and characterized, an important question has become: what are the evolutionary origins of these effectors and how did pathogens acquire them? For some filamentous plant pathogens, acquisition of effectors appears to be through the horizontal transfer of conditionally dispensable chromosomes (Hatta et al., 2002; Oliver & Solomon, 2008; Akagi et al., 2009; Ma et al., 2010). Within the Dothideomycetes, families of functionally conserved effector genes analogous to Ecp2 and Avr4 in the tomato (Lycopersicon esculentum) pathogen Cladosporium fulvum have been identified (Stergiopolous et al., 2010; Stergiopolous et al., 2012). The SnTox1 protein shares local similarity with the chitin-binding domain found in Avr4, but otherwise the proteins appear to be unrelated (Liu et al., 2012). Among more distantly related organisms, the rapid increase in genome sequences has led to the detection of HGT events involving single genes across kingdoms (Richards et al., 2011). These HGT events were identified as a result of high homology among gene sequences. With the exception of SnToxA found in P. tritici-repentis, the NEs of P. nodorum do not show homology with any known proteins in other organisms, making it impossible to reconstruct the evolutionary history of these proteins. Though all three proteins require a host gene to induce symptoms and show similar expression profiles during infection, they are not homologous with each other. Because of the absence of homologous sequences outside of P. nodorum, we relied on diversity measurements from large population samples coupled with coalescent analyses of the Phaeosphaeria species complex to infer some aspects of the evolutionary history of these genes.

As shown in Fig. 5, the last highly supported (posterior probability = 1) recent common ancestor of P. nodorum is shared with six sister species. Among these seven species, SnToxA, SnTox1 and SnTox3 were found only in P. nodorum and Pat1. The sequences of the NE genes found in Pat1 indicate that they were acquired directly from P. nodorum via hybridization, as discussed below. Taken together, our findings are consistent with at least two different evolutionary scenarios that could explain the origins of NEs within the Phaeosphaeria species complex. In one scenario, all three NEs were present in a common ancestor of the nine characterized Phaeosphaeria species and were then lost or became highly diverged in seven of these species. Under this scenario, strong selection operating on these genes generated high sequence divergence and explains our inability to detect homologous genes in the other species. We used low-stringency hybridization conditions to test this hypothesis, but could not detect homologous sequences by hybridization. Another possibility is that a presence/absence polymorphism exists for each NE in all nine species and the isolates included in the analysis were missing all three NEs. Under a second scenario, all three NE genes in P. nodorum were acquired horizontally from an unknown donor or series of donors. The geographic distribution of diversity for each NE is consistent with three separate HGT events.

McDonald et al. (2012) provided evidence of hybridization between P. nodorum and Pat1 based on sequence analysis of the β-tubulin locus. All of the effector sequence alleles found in Pat1 were also found in P. nodorum, although the two species differed significantly for conserved housekeeping genes. For SnTox3 and SnToxA, the shared alleles were also the most frequent alleles found in P. nodorum. This NE sequence data provide additional support for the hypothesis that hybridization occurred between P. nodorum and Pat1. Based on these findings we postulate that all three effectors in Pat1 were acquired from P. nodorum via inter-specific hybridization, though the extent and nature of this hybridization require further investigation.

This population genetic study is only the second to compare neutral marker diversity with diversity at effector loci to infer the evolutionary history of effector genes. The earlier population study on the barley (Hordeum vulgare) scald pathogen Rhynchosporium commune found that the highest diversity for neutral marker loci, including DNA sequences, restriction fragment length polymorphisms and microsatellites, corresponded with the highest diversity for the NIP1 effector gene in Scandinavia (Schürch et al., 2004; Brunner et al., 2007), indicating that NIP1 shared the same evolutionary history, and probably the same common ancestor, as the other R. commune genes. In contrast, the lack of geographic correspondence between toxin diversity and neutral marker diversity provides strong evidence that the NEs of P. nodorum do not share the same evolutionary history as the neutral loci.

The geographic region that harbored the highest sequence diversity was different for each effector gene. The highest diversity for SnToxA was found in South Africa, for SnTox3 in North America and Australia, and for SnTox1 in Europe. The finding of higher effector diversity in ‘New World’ populations, where wheat cultivation began only during the last few hundred years following the arrival of European colonists (South Africa c. 350 yr ago, Australia c. 200 yr ago and North America c. 500 yr ago), suggests that each of these effectors may be under strong regional selection or alternatively may have different geographic origins representing three separate HGT events. None of the populations harboring the highest effector diversity overlapped with the hypothesized center of origin of P. nodorum in the ancient Fertile Crescent. The original source of the three effector genes remains unknown, but the SnToxA protein has a domain consistent with a prokaryotic origin (Friesen et al., 2006). McDonald et al. (2012) presented evidence that P. nodorum existed as a species before the domestication of wheat, probably as a fungal endophyte on grasses. We propose that these NE genes, whether inherited from an unknown common ancestor or acquired horizontally, enabled P. nodorum to emerge from this species complex as the dominant, specialized pathogen on wheat. Hybridization with its sister species Pat1 enabled the transfer of these genes to a new fungal species, resulting in the emergence of another damaging, though closely related, pathogen.


The authors acknowledge the Genetic Diversity Center at ETH Zurich for providing sequencing facilities and the ETH Zurich for funding this research. We also thank the Australian Grains Research and Development Corporation and the USDA-NIFA AFRI Microbial Biology Program–Competitive grant no. 2009-04265 for additional funding.