The discovery of the virulence gene ToxA in the wheat and barley pathogen Bipolaris sorokiniana

Summary

Bipolaris sorokiniana is the causal agent of multiple diseases on wheat and barley and is the primary constraint to cereal production throughout South Asia. Despite its significance, the molecular basis of disease is poorly understood. To address this, the genomes of three Australian isolates of B. sorokiniana were sequenced and screened for known pathogenicity genes. Sequence analysis revealed that the isolate BRIP10943 harboured the ToxA gene, which has been associated previously with disease in the wheat pathogens Parastagonospora nodorum and Pyrenophora tritici‐repentis. Analysis of the regions flanking ToxA within B. sorokiniana revealed that it was embedded within a 12‐kb genomic element nearly identical to the corresponding regions in P. nodorum and P. tritici‐repentis. A screen of 35 Australian B. sorokiniana isolates confirmed that ToxA was present in 12 isolates. Sequencing of the ToxA genes within these isolates revealed two haplotypes, which differed by a single non‐synonymous nucleotide substitution. Pathogenicity assays showed that a B. sorokiniana isolate harbouring ToxA was more virulent on wheat lines that contained the sensitivity gene when compared with a non‐ToxA isolate. This work demonstrates that proteins that confer host‐specific virulence can be horizontally acquired across multiple species. This acquisition can dramatically increase the virulence of pathogenic strains on susceptible cultivars, which, in an agricultural setting, can have devastating economic and social impacts.

Introduction

A plant pathogen's ability to infect a host can fluctuate dramatically as a result of the presence of a single gene that confers host susceptibility or resistance (Ballance et al., 1989; Friesen et al., 2006; Meehan and Murphy, 1947). These single genes are almost exclusively pathogen specific and exert a strong qualitative effect in the facilitation of disease. One such gene is the proteinaceous necrotrophic effector ToxA. This protein was originally found in a culture filtrate of the wheat pathogenic fungus Pyrenophora tritici‐repentis, where the filtrate alone induced large necrotic lesions on susceptible wheat (Triticum aestivum) (Tuori et al., 1995). The ToxA gene and 11 kb of nearly identical surrounding DNA were later discovered in the genome sequence of a second wheat pathogen, Parastagonospora nodorum (Friesen et al., 2006). Subsequent analysis of ToxA in both species showed that P. nodorum harboured significantly more sequence diversity (Friesen et al., 2006). This finding suggested that ToxA and the surrounding region had been present in P. nodorum for a longer period of time than in P. tritici‐repentis. Additional historical evidence also showed that P. nodorum was a recognized pathogen of wheat prior to P. tritici‐repentis. Together, these data support the theory that ToxA was horizontally acquired by P. tritici‐repentis from P. nodorum (Friesen et al., 2006; Stukenbrock and McDonald, 2007). This finding has held for nearly 10 years, with very little diversity being found in ToxA within P. tritici‐repentis, but increasingly diverse haplotypes being identified in P. nodorum (Ciuffetti et al., 1997; Friesen et al., 2006; McDonald et al., 2013). ToxA was later found in a third fungal pathogen species Phaeosphaeria avenaria tritici (McDonald et al., 2012). Here, evidence of interspecific hybridization between P. nodorum and this sister species was postulated to be the mechanism of ToxA transfer (McDonald et al., 2012, 2013).

In contrast with the well‐documented examples of P. nodorum and P. tritici‐repentis, no effectors have yet been identified in the related Dothideomycete Bipolaris sorokiniana. Bipolaris sorokiniana is a pathogen of wheat and barley, and is the causal agent of several diseases, including spot blotch, Helminthosporium leaf blight (HLB) and common root rot (Duveiller et al., 2005; Knight et al., 2010). In Central Asia, HLB is a disease complex caused by both B. sorokiniana and P. tritici‐repentis, and is considered as a primary constraint to yield improvements in this food‐insecure region (Duveiller et al., 2005). Despite its significance, the diseases and causal pathogen are poorly understood. To address this knowledge gap, we sequenced the genomes of three B. sorokiniana isolates collected from either wheat or barley in Australia. These genomes were then used in a standard screen to identify any known virulence gene(s) towards wheat or barley. The unexpected discovery of the very well‐known wheat virulence gene ToxA led us to further characterize its effect on pathogenicity, its prevalence in a collection of Australian B. sorokiniana isolates and the genomic context in which it was found.

Results and Discussion

The genomes of three B. sorokiniana isolates, BRIP27492, BRIP26775 and BRIP10943, were sequenced and assembled de novo with SPAdes v3.8.0 (BioProject PRJNA349192). The assembled genome sizes were in the range 33.5–35.4 Mbp, similar in size to the reference genome B. sorokiniana ND90Pr (∼34 Mbp). BUSCO v2 was used to assess the completeness of each genome assembly using a set of core single‐copy orthologous genes (Simão et al., 2015). From this core fungal set, ∼98% of the 1438 orthologues were found in each de novo assembly. Additional genome statistics, including the N50 lengths and total number of contigs, as well as additional BUSCO statistics, are given in Table S1 (see Supporting Information).

A screen for known wheat virulence genes in these three genomes using blastn revealed that a gene with 98.1% nucleotide identity to ToxA of P. nodorum SN15 was present in the genome assembly isolate BRIP10943. To verify that the finding of ToxA in BRIP10943 was not a result of contamination in the DNA extraction, the entire de novo genome assembly was employed to construct the phylogenetic distance tree shown in Fig. 1 using andi (Haubold et al., 2015). Isolate BRIP10943 clearly resolved with the two other sequenced B. sorokiniana genomes (this study) and the reference genome isolate ND90Pr (Condon et al., 2013). Figure 1 also shows the correct relationship between B. sorokiniana and its sister species B. maydis and B. victoriae (Schoch et al., 2009). These data confirm that BRIP10943 is a pure culture of B. sorokiniana that possesses the necrotrophic effector ToxA.

Figure 1.

Phylogenetic relationship amongst plant pathogen Pleosporales, including Parastagonospora nodorum SN15 and Pyrenophora tritici‐repentis 1C‐BFP. The ToxA‐containing strain of Bipolaris sorokiniana is labelled with a red asterisk. Numbers on the branches show the percentage support from 1000 bootstrap replicates and the tree is drawn with a midpoint root.

To test whether ToxA functions in a Tsn1‐dependent manner in B. sorokiniana, three cultivars of wheat, BG261 (Tsn1), Espada (tsn1) and Westonia (tsn1), were inoculated with an isolate of B. sorokiniana containing ToxA (BRIP10943) and another isolate lacking ToxA (BRIP27492). These pathogenicity assays showed that the isolate carrying ToxA was more virulent on wheat seedlings carrying the sensitivity gene Tsn1 (Fig. 2). In contrast, both isolates appeared to be equally virulent on the cultivar Espada, whereas BRIP27492 (–ToxA) appeared to be slightly more virulent on the cultivar Westonia. These two isolates also appeared to be equally virulent on the secondary host barley (Fig. S1, see Supporting Information). This result demonstrates that the disease is more severe on a Tsn1‐containing wheat cultivar in the presence of ToxA. It also shows that there are other wheat pathogenicity genes that remain unknown, as the isolate lacking ToxA was still highly virulent on non‐Tsn1 cultivars.

Figure 2.

Whole‐seedling pathogenicity assays of Bipolaris sorokiniana isolates BRIP10943 (ToxA) and BRIP27492 (toxA) on the wheat cultivars BG261 (Tsn1), Espada (tsn1) and Westonia (tsn1). Symptom development on BG261 infected with the BRIP10943 isolate was significantly more severe compared with the BRIP27492 isolate. In contrast, symptoms caused by each of the isolates were comparable on the assayed tsn1 wheat lines, demonstrating that ToxA confers a fitness advantage on Tsn1 wheat lines.

ToxA was found on a 15.9‐kb scaffold within the de novo genome assembly of isolate BRIP10943. This scaffold was aligned with the ToxA‐containing 2.7‐Mbp chromosome from P. tritici‐repentis 1C‐BFP and the 50‐kb scaffold from P. nodorum SN15 using progressive Mauve. The total length of the longest collinear block identified by Mauve covered 68 kb on the P. tritici‐repentis 1C‐BFP chromosome with large gaps between AT‐rich regions (Fig. S2, see Supporting Information). Between the three species there is a region of approximately 12 kb in length, shown in Fig. 3A, with an average pairwise identity of 91.9%. This 12‐kb region corresponds exactly to the originally reported region that was deemed to be horizontally transferred from P. nodorum to P. tritici‐repentis (Friesen et al., 2006). The strong pair‐wise identity across such a large DNA fragment clearly suggests that this region was horizontally transferred.

Figure 3.

Alignment of the ToxA‐containing region from Pyrenophora tritici‐repentis, Parastagonospora nodorum and Bipolaris sorokiniana. (A) Total alignment of the 12‐kb homologous region between the three species. This view shows the differences in AT richness (green line); large increases are denoted by red asterisks. The decay in identity near the edges of the region can be attributed to repeat‐induced polymorphism (RIP). All annotated genes are shown using the gene accession number from P. nodorum (green bars). (B) Differences in the promoter of ToxA as a result of the presence of small indels. In P. tritici‐repentis, there is a 238‐bp insertion, which has no homology to any other organism. In P. nodorum, a single 43‐bp indel is present, which is a simple repeat of the same 43 bp preceding the indel (blue bars). In B. sorokiniana, there is a 148‐bp indel, which forms a near‐perfect DNA hairpin.

Despite the high degree of homology across the 12‐kb region, there are some notable differences between the three species. Annotation of the aligned ToxA promoter revealed a short 43‐bp indel in P. tritici‐repentis and B. sorokiniana, which is 517 bp upstream of the translation start site. Figure 3B shows that this 43‐bp indel is caused by a short tandem repeat that is present in P. nodorum twice, but only once in the other two species (denoted TR_1 and TR_2 in Fig. 3B). Another 43 bp upstream of this repeat is a 141‐bp insertion that is unique to B. sorokiniana BRIP10943 (Fig. 3A). Annotation of this unique sequence revealed that this indel forms a perfect DNA hairpin (Fig. S3, see Supporting Information). The discovery of this hairpin in the promoter of B. sorokiniana led to a broader characterization of other inverted repeats (IRs) in the region, shown in Fig. S2. We identified one other near‐perfect hairpin in B. sorokiniana, ∼5.3 kb upstream of the ToxA start codon (Fig. 3B). In fungi, there is little literature describing DNA hairpins. In prokaryotes and phages, however, these DNA structures have been shown to act as origins of replication and as binding sites for transcription factors and other regulatory proteins (Bikard et al., 2010).

As the sequence similarity decays near the edges of the conserved 12‐kb region, there is a coincident increase in the AT richness of the sequence. The specific points at which the AT richness increases are denoted by red asterisks in Fig. 3A. This is caused mainly by the activity of repeat‐induced polymorphism (RIP) in P. nodorum and P. tritici‐repentis. RIP is a well‐described fungal genome defence system that protects against self‐replicating repetitive elements (Selker, 1990). During meiosis, the RIP machinery identifies invasive repetitive DNA and introduces CpG to CpA mutations. This leads to the introduction of stop codons in coding regions, thereby preventing the transposon machinery from operating (Hane and Oliver, 2008; Selker, 1990). In P. tritici‐repentis, RIP begins approximately 3.5 kb upstream of the ToxA start codon. The activity of RIP in P. tritici‐repentis led to the introduction of early stop codons in SNOG_16569 and SNOG_16570 (Fig. 3B). In P. nodorum, this pattern does not begin until 5 kb upstream of the ToxA start codon and these two genes remain intact (Fig. 3). Based on the AT richness plot and, in comparison with the other two species, there is no evidence of RIP in any of the aligned regions in B. sorokiniana (Fig. 3).

In a large comparison of Dothideomycete genomes, all of the machinery for RIP was annotated and present in all three species (Ohm et al., 2012). To investigate whether the absence of RIP on the ToxA scaffold in BRIP10943 was caused by a mutation in the RIP machinery, we identified long terminal repeat (LTR) transposases within the entire de novo genome assembly. Using two of these predicted transposases, we identified other highly similar sequences within the genome using local blastn. The top 100 blast hits were aligned and their RIP index was measured. Both LTR transposase families showed strong evidence of RIP in the genome of BRIP10943 (Fig. S4, see Supporting Information). This clearly shows that the mechanisms for RIP are intact and active in BRIP10943. In P. nodorum and P. tritici‐repentis, the presence of RIP on the ToxA scaffold was attributed to the activity of the transposable elements found within the 12‐kb region (Friesen et al., 2006). The absence of RIP in the 12 kb of B. sorokiniana implies that the donor organism or donor DNA possessed an un‐RIPed copy of this region. Further, in B. sorokiniana, the repetitive elements within this 12 kb have not yet attracted the RIP machinery. We hypothesize that this is either a result of their inactivity or, alternatively, that RIP has not yet targeted these transposases because the horizontal transfer is more recent than that of P. nodorum and P. tritici‐repentis.

Figure 4.

(A) Haplotype network showing the distribution of ToxA sequence diversity across species. Each circle represents a unique ToxA sequence and the colour corresponds to the species as denoted in the legend. The short black hashed lines show the number of single nucleotide polymorphisms between each haplotype. Seven of the 12 Bipolaris sorokiniana isolates have a nucleotide sequence that is identical to that found in Pyrenophora tritici‐repentis. The remaining five isolates have a single non‐synonymous substitution. Other published ToxA sequences are only drawn once as a single isolate in this network. Two previously published haplotypes collapsed into a single circle as a result of a slightly shorter sequence used for alignment in this article. (B) Unique amino acid sequence alignment for all ToxA haplotypes and the unique sequence from this study. Numbers beside the name in parentheses indicate the number of nucleotide haplotypes contained within that amino acid haplotype.

To better understand the prevalence and diversity of ToxA in B. sorokiniana, we examined 35 isolates from different field locations in Australia (Table S2, see Supporting Information). For these isolates, we confirmed their classification as B. sorokiniana by amplifying a short segment of elongation factor‐1alpha (Ef‐1α, ND90Pr Transcript ID:EMD60564; Carbone and Kohn, 1999). We compared this region with the orthologous gene in the sister species B. victoriae FI3. Between B. victoriae FI3 and the collected B. sorokiniana isolates, there were 12 fixed single nucleotide polymorphisms (SNPs) (Fig. S5, see Supporting Information). These data show that all the isolates in this study for which we do not have whole‐genome data are B. sorokiniana.

Using a polymerase chain reaction (PCR) assay, we found ToxA in 12 of the 35 B. sorokiniana isolates (34.2%). These 12 isolates were randomly distributed across all three Australian states in which samples were collected (Table S2). Most locations are only represented by one isolate in our collection, and so we cannot infer the prevalence of ToxA in these field populations. Our earliest sample that contains ToxA is isolate BRIP10943, which was collected from wheat in 1966. Of the 12 ToxA‐positive isolates, 11 were collected from wheat (Triticum aestivum) and one from barley (Hordeum vulgare). We are unable to draw any further conclusions about the frequency of ToxA in wheat‐ or barley‐infecting populations as we lack sufficient population‐level sampling from either host. In P. nodorum and P. tritici‐repentis, ToxA is known to persist as a presence/absence polymorphism (Friesen et al., 2006). In P. nodorum, the presence or absence of ToxA in a global population dataset was found to vary drastically depending on the population sampled (McDonald et al., 2013). For Australian populations of P. nodorum and P. tritici‐repentis, nearly 100% of the tested isolates contained ToxA (Antoni et al., 2010; McDonald et al., 2013). The high prevalence of ToxA in Australian populations of P. nodorum and P. tritici‐repentis was attributed to a founding population of isolates that all possessed ToxA. The proportion of ToxA‐positive isolates in B. sorokiniana observed in this study (34%), however, is dramatically different compared with the near 100% presence reported in the other two species. Further sampling of structured field populations is required to determine the true prevalence of ToxA in Australian B. sorokiniana populations.

We obtained sequence data from all 12 ToxA‐positive isolates, which revealed two B. sorokiniana nucleotide haplotypes (Fig. 4A). One haplotype matched exactly the only publically available ToxA sequence for P. tritici‐repentis (Fig. 4A). The other haplotype, which differed by a single base pair, conferred a non‐synonymous substitution that was identical to P. nodorum (Fig. 4B). There is sufficient evidence to suggest that the ToxA regions in B. sorokiniana and P. tritici‐repentis share a closer evolutionary origin with each other than they do with P. nodorum. This is based on the identical ToxA sequence shared between P. tritici‐repentis ToxA and one haplotype of B. sorokiniana (Fig. 4A). In Central Asia, HLB is described as a disease complex consisting of both B. sorokiniana and P. tritici‐repentis (Duveiller et al., 2005). This close association clearly provides the opportunity for the two pathogens to encounter each other in the field and potentially exchange DNA. Sampling of populations of both P. tritici‐repentis and B. sorokiniana from the regions in Central Asia in which spot blotch is a significant disease will greatly aid in answering this question. Although the exact identity of the original donor remains obscure, the substantial homology across the 12‐kb genomic region in which ToxA is embedded clearly indicates a common evolutionary origin for all three fungal species.

ToxA has been demonstrated previously to play a dominant role in the wheat diseases caused by P. nodorum and P. tritici‐repentis (Liu et al., 2006). In this study, we have demonstrated that this influential pathogenicity protein has now been discovered in the emerging wheat pathogen B. sorokiniana. This is significant for two reasons. First, these data demonstrate that ToxA is embedded in a highly mobile ‘pathogenicity element’, which facilitates disease on wheat in three globally important fungal pathogens. Second, and of a more practical nature, our intimate knowledge of this gene‐for‐gene interaction will allow us to make immediate changes to disease management strategies. This is especially important for the food‐insecure regions of wheat in Central Asia, where this disease is reported as a major yield threat (Sharma and Duveiller, 2006). This threat is likely to increase in the face of climate change (Sharma et al., 2007). A very clear first step to mitigate losses is to screen for the prevalence of Tsn1 in Asian wheat cultivars. This strategy has been successful in limiting the impact of P. nodorum and P. tritici‐repentis in Australia (Oliver et al., 2009). Although the evolutionary origins of this gene and the mechanisms facilitating its transfer into multiple wheat pathogens remain intriguing open questions, prevention of losses via the removal of Tsn1 from cultivated wheat in disease‐conducive environments, such as Central Asia, will have a major positive effect on global food security.

Experimental Procedures

Details of the sampling dates and locations for each isolate are listed in Table S2. All fungal cultures were grown on potato dextrose agar (PDA, BD, Franklin Lakes, USA) at 22°C under a 12‐h light/dark cycle. For long‐term storage, fungal spores scraped from PDA plates were frozen at −80°C in 25% glycerol. DNA was extracted from freeze‐dried fungal tissue by grinding in 1% SDS TE (1% sodium dodecyl sulphate, 10 mM Tris, 10 mM Na‐EDTA, pH 8.0) buffer, followed by ethanol precipitation. PCRs were performed with published ToxA and Ef‐1α primers, using TakaraTaq (Scientifix, Cheltenham, Australia) (Carbone and Kohn, 1999; Friesen et al., 2006). Sanger and Illumina paired‐end sequencing was performed at the Biomolecular Resource Facility at The Australian National University.

Whole‐seedling pathogenicity assays were conducted on 2‐week‐old plants, grown in a controlled environment chamber under a 14 day/night cycle (20°C day, 12°C night) at 85% relative humidity. Fungal spores were prepared from PDA plates and diluted to a concentration of 1 × 10⁴ spores/mL in 0.02% Tween‐20. Inoculated seedlings were incubated at 20°C in the dark at 100% humidity for 48 h before being transferred to a controlled environment chamber. Photographs were taken at 5 days post‐inoculation. For detached leaf pathogenicity assays, the inoculum was prepared as described above and sprayed onto detached wheat and barley leaves of 5 cm in length. After spraying, leaves were embedded in 0.8% tap water agar that contained 150 mg/L benzimidazole. Photographs were taken at 4 days post‐inoculation.

De novo genome assembly was performed with SPAdes v3.8.0 (Bankevich et al., 2012). Scaffolds/chromosomes that contained ToxA were aligned using Progressive Mauve v2.3.1 as implemented in Geneious v9.1.4 (Darling et al., 2010). The whole‐genome phylogenetic distance tree and 1000 bootstrap replicates were calculated with andi v0.9.3 (Haubold et al., 2015; Klötzl and Haubold, 2016). The resulting 1000 distance matrices were clustered using PhyD* employing the BioNJ* method and no outgroup (Criscuolo and Gascuel, 2008). The resulting Newick tree was plotted using FigTree v1.4.0 and drawn with a midpoint root (http://tree.bio.ed.ac.uk/software/figtree/).

To examine the homology on the ToxA‐containing scaffolds alone, an initial Mauve alignment was conducted with the full‐length scaffold from P. tritici‐repentis (http://fungi.ensembl.org/Pyrenophora_triticirepentis/Info/Index) and P. nodorum (http://fungi.ensembl.org/Pyrenophora_triticirepentis/Info/Index) (Manning et al., 2013; Syme et al., 2016) and the de novo scaffold from B. sorokiniana. The Mauve alignment was conducted with the setting to automatically calculate the seed weight and the minimum Locally Collinear Block (LCB) score. After the first Mauve alignment, the sequences within the LCB containing ToxA from all three species were extracted and re‐aligned with the same settings. Other DNA alignments were performed using Muscle v3.5 implemented as a plugin in Geneious (Edgar, 2004). The IRs were identified using Emboss's einverted as implemented at http://emboss.bioinformatics.nl/cgi-bin/emboss/einverted with the following settings: gap penalty, 12; min score threshold, 50; match score, 3; mismatch score, −4; max length of repeats, 2000. The long interspersed repeat transposons were identified with LTRharvest with the following options: ‐minlenltr 250 ‐maxlenltr 5000 ‐similar 75 (Ellinghaus et al., 2008). Local blastn was performed with an e‐value cut‐off of 1 × 10⁻³ and limited to a maximum of 100 hits (Camacho et al., 2009).

Supporting information

Additional Supporting Information may be found in the online version of this article at the publisher's website:

Table S1 Genome assembly and quality statistics.

Table S2 Bipolaris sorokiniana isolate details, genome accession numbers and ToxA haplotype accession numbers.

Fig. S1 Detached leaf pathogenicity assays with Bipolaris sorokiniana isolates BRIP10943 (harbouring ToxA) and BRIP27492 (lacking ToxA) on barley and wheat. Disease symptoms for both isolates were comparable on barley, indicating that ToxA does not provide a fitness advantage on this host. By comparison, disease symptoms on Tsn1‐containing wheat were consistent with the seedling assays presented in Fig. 2, showing that the ToxA‐harbouring isolate was more virulent.

Fig. S2 Mauve alignment of the larger ToxA homology between Pyrenophora tritici‐repentis and Parastagonospora nodorum. Green bars indicate annotated genes from P. nodorum, whereas blue bars indicate annotated genes from P. tritici‐repentis. Orange bars show inverted repeats or other repetitive elements. The AT richness plot is shown with the green line at the bottom of the plot. Large increases in AT richness are denoted with red asterisks.

Fig. S3 Alignment of the DNA hairpin found in the promoter of ToxA. The top sequence is the complete hairpin sequence and the bottom sequence is the reverse complement of the same sequence.

Fig. S4 RipCal graph of two long terminal repeat (LTR) transposon families identified in Bipolaris sorokiniana isolate CS10943. The top sequence was identified as the most GC‐rich sequence in the alignment and all other sequences were analysed for CpG to CpA mutations based on the GC‐rich sequence. The high repeat‐induced polymorphism (RIP) index in both of these families is evidence that the RIP machinery is present and active in isolate CS10943.

Fig. S5 Alignment of elongation factor‐1alpha (Ef‐1α) for all isolates tested for the presence of ToxA via polymerase chain reaction (PCR). Reference genomes Bipolaris maydis C5 and Bipolaris victoriae F13 were downloaded from Ensembl Fungi (fungi.ensembl.org). Single nucleotide polymorphisms (SNPs) are noted between B. sorokiniana and its closest relative B. victoriae only. This alignment shows that all strains used in this study are B. sorokiniana.