Variability in an effector gene promoter of a necrotrophic fungal pathogen dictates epistasis and effector-triggered susceptibility in wheat

The fungus Parastagonospora nodorum uses proteinaceous necrotrophic effectors (NEs) to induce tissue necrosis on wheat leaves during infection, leading to the symptoms of septoria nodorum blotch (SNB). The NEs Tox1 and Tox3 induce necrosis on wheat possessing the dominant susceptibility genes Snn1 and Snn3B1/Snn3D1, respectively. We previously observed that Tox1 is epistatic to the expression of Tox3 and a quantitative trait locus (QTL) on chromosome 2A that contributes to SNB resistance/susceptibility. The expression of Tox1 is significantly higher in the Australian strain SN15 compared to the American strain SN4. Inspection of the Tox1 promoter region revealed a 401 bp promoter genetic element in SN4 positioned 267 bp upstream of the start codon that is absent in SN15, called PE401. Analysis of the world-wide P. nodorum population revealed that a high proportion of Northern Hemisphere isolates possess PE401 whereas the opposite was observed in representative P. nodorum isolates from Australia and South Africa. The presence of PE401 removed the epistatic effect of Tox1 on the contribution of the SNB 2A QTL but not Tox3. PE401 was introduced into the Tox1 promoter regulatory region in SN15 to test for direct regulatory roles. Tox1 expression was markedly reduced in the presence of PE401. This suggests a repressor molecule(s) binds PE401 and inhibits Tox1 transcription. Infection assays also demonstrated that P. nodorum which lacks PE401 is more pathogenic on Snn1 wheat varieties than P. nodorum carrying PE401. An infection competition assay between P. nodorum isogenic strains with and without PE401 indicated that the higher Tox1-expressing strain rescued the reduced virulence of the lower Tox1-expressing strain on Snn1 wheat. Our study demonstrated that Tox1 exhibits both ‘selfish’ and ‘altruistic’ characteristics. This offers an insight into a complex NE-NE interaction that is occurring within the P. nodorum population. The importance of PE401 in breeding for SNB resistance in wheat is discussed.


Abstract
The fungus Parastagonospora nodorum uses proteinaceous necrotrophic effectors (NEs) to induce tissue necrosis on wheat leaves during infection, leading to the symptoms of septoria nodorum blotch (SNB). The NEs Tox1 and Tox3 induce necrosis on wheat possessing the dominant susceptibility genes Snn1 and Snn3B1/Snn3D1, respectively. We previously observed that Tox1 is epistatic to the expression of Tox3 and a quantitative trait locus (QTL) on chromosome 2A that contributes to SNB resistance/susceptibility. The expression of Tox1 is significantly higher in the Australian strain SN15 compared to the American strain SN4. Inspection of the Tox1 promoter region revealed a 401 bp promoter genetic element in SN4 positioned 267 bp upstream of the start codon that is absent in SN15, called PE401. Analysis of the world-wide P. nodorum population revealed that a high proportion of Northern Hemisphere isolates possess PE401 whereas the opposite was observed in representative P. nodorum isolates from Australia and South Africa. The presence of PE401 removed the epistatic effect of Tox1 on the contribution of the SNB 2A QTL but not Tox3. PE401 was introduced into the Tox1 promoter regulatory region in SN15 to test for direct regulatory roles. Tox1 expression was markedly reduced in the presence of PE401. This suggests a repressor molecule(s) binds PE401 and inhibits Tox1 transcription. Infection assays also demonstrated that P. nodorum which lacks PE401 is more pathogenic on Snn1 wheat varieties than P. nodorum carrying PE401. An infection competition assay between P. nodorum isogenic strains with and without PE401 indicated that the higher Tox1-expressing strain rescued the reduced virulence of the lower Tox1-expressing strain on Snn1 wheat. Our study demonstrated that Tox1 exhibits both 'selfish' and 'altruistic' characteristics. This

Introduction
Breeding for effective resistance to fungal diseases in crops is a continual challenge. Fungal pathogens, namely biotrophs and hemibiotrophs, evolve ways to overcome host resistance by masking themselves to evade pathogen-triggered immunity. This can occur through mutations in Avr-effector genes, leading to the loss of recognition and the subsequent breakdown of host resistance that ultimately results in 'boom and bust' cycles [1]. This is manifested through the mass adoption of a resistant cultivar (boom). The pathogen population exposed to the resistant gene will then adapt by selecting favourable mutations in the Avr gene that leads to a loss of host recognition and resistance (bust). Notable examples occur in the rust and blackleg pathosystems [2,3]. In contrast, necrotrophic fungal pathogens belonging to the Pleosporales such as the tan spot of wheat fungus Pyrenophora tritici-repentis (Ptr) and the septoria nodorum blotch (SNB) of wheat fungus Parastanonospora nodorum secrete host-specific necrotrophic effectors (NEs) to cause tissue necrosis on host plants possessing a matching dominant susceptibility gene. This results in effector-triggered susceptibility, enabling the pathogen to proliferate [4,5]. Improvements in host resistance to necrotrophic fungal pathogens rely on the removal of the host dominant-susceptibility genes [6]. However, necrotrophic fungal pathogens often possess multiple NEs that exploit different dominant-susceptibility genes carried by the host [7]. Hence, these functional redundancies often require breeders to remove multiple host genes or stack the required number of desirable disease resistance alleles to be effective [8,9]. These situations make it difficult for breeders to provide long-lasting and durable resistance. P. nodorum utilises proteinaceous necrotrophic effectors (NE)s to cause necrosis on host plants with a susceptible genotype, resulting in necrosis that facilitates host infection [10][11][12]. Genetic evidence has revealed at least 10 NE-dominant susceptibility gene interactions in the P. nodorum-wheat pathosystem. These are ToxA-Tsn1 [13,14], Tox1-Snn1 [15,16] [21] have been cloned and characterised for their roles in virulence. These NEs are secreted, generally have a high cysteine content and are relatively small in molecular mass (<30kDa). Of the host dominant-susceptibility receptors, the genes for Tsn1 [14], Snn1 [15] and Snn3D1 [19] have been cloned. These genes encode proteins similar to canonical nucleotide-binding leucine-rich repeat resistance gene receptors that typically possess a protein kinase domain coupled with nucleotide-binding leucine-rich repeats or major sperm protein domains. Under controlled settings, the removal of a NE or the host receptor from the pathosystem results in the loss of ETS and a drastic reduction of P. nodorum virulence on wheat [eg. 13,16,18].
SNB resistance is quantitatively inherited [23] and improvements in resistance have been, in part, mediated through the removal of dominant susceptibility genes in the host [6]. However, the situation is complex. Multiple NEs expressed by P. nodorum afford the pathogen functional redundancies on wheat varieties that possess multiple dominant susceptibility genes [8,24]. For instance, Tan et al. [8] demonstrated that the deletion of ToxA, Tox1 and Tox3 in P. nodorum SN15 did not significantly alter virulence on a diverse collection of modern wheat varieties [8,9]. A further complication in breeding for SNB resistance is that some NE-host gene interactions were not consistently detected in field trials using mapping populations with known susceptibility genes [25,26]. It is intuitive to hypothesise that NE epistasis plays a significant role in shaping the variations observed in host resistance to SNB (Table 1) [27]. NE epistasis can be defined as interactions between NE genes where the effect conferred by one is masked by the presence of another [27]. NE epistasis in the P. nodorum-wheat pathosystem can be divided into two broad regulatory categories. Firstly, the contribution of a NE-host dominant susceptibility gene interaction is suppressed by another competing NE-host gene interaction possibly through host-gene action or cross-talk among NE recognition pathways [eg. 28,29]. Secondly, NE gene repression mediated by the expression of another NE [eg. 22,30]. For the latter, a clear-cut example involves the suppression of the Tox3-Snn3 interaction in SNB by Tox1-Snn1 described by Phan et al. [22]. This was explored as part of a study that mapped SNB on a double-haploid (DH) wheat population derived from two major Australian commercial wheat cultivars. SNB quantitative trait loci (QTL) detected during infection by the Australian P. nodorum reference isolate SN15 were compared with a Tox1 deletion mutant (tox1-6) [22]. This revealed the Tox1-Snn1 NE-host receptor interaction was epistatic to Tox3-Snn3B1 and a major SNB QTL located on chromosome 2A in the DH population that

PLOS PATHOGENS
A mechanism of effector epistasis confers sensitivity to a novel NE which we coined as Tox2A. Quantitative RT-PCR revealed that Tox3 expression was significantly higher in tox1-6 than in the wildtype, which presented a basis for NE epistasis mediated by gene repression [22]. The expression of Tox1 also varies among isolates. As Tox1 production was much higher in SN2000 than both SN4 and SN6, the former was used to enable its detection and functional characterisation [16,32,33]. This contrasts with other USA isolates such as SN4 and SN6 where Tox1 expression is low relative to SN2000 [34]. A recent study also demonstrated that the Australian P. nodorum SN15 isolate expressed Tox1 significantly higher than SN4, which correlated with a greater SNB disease contribution mediated by the Tox1-Snn1 interaction [31]. The mechanisms behind this differential expression remain unexplored however, it was noted that the Tox1 promoter region was found to be polymorphic between these isolates [31]. These observations prompted us to ask several key questions. Firstly, does the genetic polymorphism in the promoter region drive differential Tox1 expression and affect epistatic outcome of Tox3-Snn3B1 and Tox2A-Qsnb.cur-2AS1 interactions. Secondly, what is the possible mechanism of Tox1 regulation? Thirdly, is there a fitness penalty associated with lower Tox1 expression in P. nodorum isolates? And lastly, what is the international distribution of Tox1-expressing P. nodorum variants? The study presented here was undertaken to characterise the gene regulatory elements controlling Tox1 expression in P. nodorum and explore their role in NE epistasis with end-goals of 1. improving SNB management strategies in wheat growing regions where the disease is prominent and 2. to provide an insight into a mechanism of NE epistasis that may dictate the outcome of fungalhost plant pathosystems.

Genetic variability in the Tox1 promoter region
To determine why both Tox1-mediated epistasis and Tox1 expression differed between P. nodorum isolates, we aligned Tox1 promoters sequences from a geographically diverse set of 20 P. nodorum genomes [35-37] (Fig 1A). The alignment revealed a 401 bp genetic element that was absent in four isolates within the collection. This element, henceforth called PE401 was absent in the Australian P. nodorum SN15 and SN2000 isolates that were associated with a higher level of Tox1 expression ( Fig 1A) [16,31,34]. The size and proximity to the gene within the Tox1 promoter suggested PE401 was the most plausible variant involved in gene repression, despite the presence of other smaller genetic polymorphisms.
PE401 was then examined for features of a mobile genetic element (MGE), as they are known to modulate the expression of proximal fungal genes [38][39][40][41]. The SN4 promoter sequence was queried against the Repbase and Rfam databases. No matches to characterised MGEs or expressed non-coding elements were observed. BLAST analysis of PE401 identified a single region with 63% identity in the SN15 genome, partially overlapping the annotated gene SNOG_30065 [35]. However, the SNOG_30065 predicted peptide sequence did not contain annotated protein domains, nor was the gene expressed under in vitro conditions and during host infection [42]. Therefore, SNOG_30065 is likely a pseudogene. No further similar matches to PE401 were detected when queried against the NCBI nucleotide and whole-genome shotgun collections, suggesting the sequence was not a repetitive element. Terminal repeats characteristic of fungal MGEs were not identified, although the GC % was low (41.9%) relative to the rest of the SN4 promoter region (48.9%) [43] (Fig 1B and  1C). We concluded that PE401 is unlikely to be a transposable element or functional gene in P. nodorum.

PLOS PATHOGENS
A mechanism of effector epistasis PLOS Pathogens | https://doi.org/10.1371/journal.ppat.1010149 January 6, 2022 Evidence of regional distribution for PE401 in the international P. nodorum population The distribution of PE401 was determined in a international panel of 484 Tox1-containing P. nodorum isolates. It was observed that isolates carrying PE401 were ubiquitous in Europe (Switzerland, Denmark, Finland, Netherlands and Norway), most sampling sites in North America (USA and Canada) and also in Iran. The latter region was previously proposed as a centre of origin for P. nodorum NEs including Tox1 [44,45]. Isolates that lacked PE401 were dominant in Australia (87% of 180 isolates) and South Africa (100% of 22 isolates), while 27% of all isolates from North Dakota (USA) and a single isolate from Texas (USA) also lacked PE401 (Fig 2). A. An alignment highlighting promoter variants in a geographically diverse selection of 20 Tox1-containing P. nodorum isolates. Within the 1,500 bp promoter region, the 401 bp PE401 and a 14 bp GTTTTCGGCCGTAT tandem repeat polymorphism are indicated. B. A GC% plot using a 50 bp average sliding window covering the Tox1 promoter from the USA reference isolate SN4. PE401 has a low GC% (blue) relative to directly adjacent promoter regions. AT% (green) is also indicated. C. A dot plot of PE401 using a 5 bp sliding window demonstrating the paucity of tandem/inverted repeat stretches. Genome coverage for of the promoter region of Tox1 in NOR-4 is incomplete. https://doi.org/10.1371/journal.ppat.1010149.g001

PLOS PATHOGENS
A mechanism of effector epistasis

PE401 functions as a repressor of Tox1 expression
We hypothesised that PE401 dictates Tox1 expression based on the differential expression profile between SN4 (+PE401) and SN15 (-PE401) carrying the two versions of the promoter [31]. Therefore, Tox1 expression was assessed in SN4, SN15 and three representative Australian P. nodorum wildtype isolates with or without PE401 (Fig 3A). Quantitative RT-PCR analysis revealed the two Australian isolates carrying PE401, WAC13443 and WAC13072, shared a low Tox1 expression profile comparable to SN4. On the other hand, Tox1 expression was significantly higher in SN15 and WAC13690 which lacked PE401 (Fig 3B). Sequence alignment of the promoter region revealed WAC13443 and WAC13072 does not carry the 14 bp tandem repeat found in SN4, suggesting that PE401 was the variant responsible ( Fig 3A).
We then created isogenic strains with and without PE401 to eliminate the possibility of strain-specific genetic factors located outside of the Tox1 locus involved in gene repression. The constructs used for promoter replacement included 873 bp of the Tox1 promoter either with PE401 (+), without PE401 (-) or with a 401 bp spacer (S) (Fig 3C). These were independently transformed into SN15 and WAC13443 at the Tox1 locus through homologous recombination. Quantitative RT-PCR analysis of the mutants revealed a 40-fold (SN15) and 38-fold (WAC13443) reduction in Tox1 gene expression occurred in the presence of PE401 when grown in Fries3 broth in vitro. Tox1 was not repressed in mutants that carry the spacer sequence, suggesting the effect was attributable to sequence specificity conferred by PE401 ( Fig 3D). We concluded that PE401 is associated with Tox1 repression, perhaps through an interaction with a trans-acting regulatory factor.
We then determined the regulatory role of PE401 in Tox1 expression during host infection using qRT-PCR. Infection time points were chosen to represent major stages of infection such as penetration (3 dpi), colonisation (5 dpi) and sporulation (7 dpi) [46] (Fig 4). As expected, Tox1 expression was much lower (log power) in SN15 and WAC13443 mutants carrying PE401 than mutants carrying either the spacer sequence or lacking PE401 throughout the entire infection period (Fig 4).
It was previously observed that deletion of Tox1 in SN15 caused a significant increase in Tox3 expression [22]. Therefore, we hypothesised that the inclusion of PE401 would alleviate

PLOS PATHOGENS
A mechanism of effector epistasis

PLOS PATHOGENS
A mechanism of effector epistasis Tox3 repression by Tox1. However, the decrease in Tox1 expression conferred by PE401 did not affect Tox3 expression (Fig 4). This indicated Tox1 repression caused by PE401 was insufficient to alleviate epistasis of Tox3 and that full suppression of Tox1 expression is needed to achieve this.

PE401 influences Tox1-Snn1-mediated ETS in wheat
The contribution of Tox1-Snn1 to SNB symptoms caused by SN15 (+) and SN15 (-) mutant strains were determined using an association mapping approach. We used a Calingiri (Snn1, snn3b1, tsn1) and Wyalkatchem (snn1, Snn3B1, tsn1) (CxW) DH wheat population infected with mutant spores at the seedling stage to directly compare the effect of PE401 repression on Tox1-Snn1 and other important SNB interactions. SNB seedling resistance QTL can be detected in the CxW population on chromosomes 1BS (Snn1), 5BS (Snn3B1), 2A, 2D, 3A and 4B when infected with SN15 or its derivative NE deletion mutants [22]. The seedling infection assay revealed the average disease score on the 179 wheat lines (including parental lines) for the SN15 (+) mutant was 4.4. This is significantly lower than SN15 (-) at 4.8 ( Table 2). We then compared the SNB severity of SN15 (+) and (-) on genotype combinations that possess Snn1 and observed that the former isolates were significantly more pathogenic ( Table 2). In contrast, no significant difference was observed between SN15 (-) and SN15 (+) on snn1 wheat lines.
There was no difference in disease severity on Snn3 versus snn3 wheat for either mutant, which had previously been observed for the tox1 mutant compared to SN15 [22]. This indicated that Tox1 remained epistatic to Tox3 when expressed at the lower level conferred by the PE401. Reliable markers for the 2A, 2D, 3A and 4B resistance loci are not yet available which meant the disease severity could not be directly compared between the SN15 (-) and (+) mutants. Nevertheless, the results suggest that the Tox1-Snn1 interaction contributed significantly to SNB severity, and that higher expression of Tox1 conferred by the absence of PE401 produced a significantly higher level of disease on Snn1 wheat.

The absence of PE401 results in higher biomass on Snn1 wheat lines
We hypothesised that P. nodorum isolates lacking PE401 resulting in higher Tox1 expression are fitter on Snn1 wheats. However, low Tox1-expressing P. nodorum isolates that possess PE401 are maintained in the population pool at a low frequency, S1 Fig and [47]. To determine evidence of fitness penalty on Snn1 wheat, we used a digital PCR approach to compare the DNA biomass of SN15 (+) and (-) during infection of Snn1 (H086, H336 and H091) and snn1 (H213 and H324) wheat varieties deriving from the CxW population that lacked Tox3 and Tox2A sensitivity ( Fig 5). Fungal biomass was determined at two and four dpi where Tox1 is Table 2. Seedling SNB rating of the CxW DH population grouped by Tox1 and Tox3 sensitivity conferred by Snn1 and Snn3B1, respectively. SN15 isogenic mutants with (+) and without (-) PE401 were used to assess seedling wheat infection. Mean SNB scores and SD were determined for both promoter replacement mutants. A paired Student's T-test (n = 3) was used to compare the mean between strains for each wheat genotype groups. � Indicates significant differences between the mutants at the P<0.05 significance threshold. Disease scores for all lines are provided in S3 File.

PLOS PATHOGENS
A mechanism of effector epistasis maximally expressed. For snn1 wheat varieties infected with either SN15 (+) or (-), there was no significant difference (P > 0.05) between the biomass of SN15 (+) and (-) at two or four dpi, with an expected increase of biomass over time. For wheat varieties with Snn1, the biomass of SN15 (-) was significantly higher than SN15 (+) at two and four dpi ( Fig 5). We then co-infected all five wheat varieties with an equal ratio of SN15 (+) and (-) to determine if SN15 (-) would outcompete SN15 (+) during infection on Snn1 wheats using a basic De Wit replacement series of 50:50 input replacement combining SN15 (-) and (+) [48]. We observed the proportion of SN15 (+) and (-) biomass was near-identical at both time points in contrast to the initial hypothesis. This indicates that SN15 (-) assisted SN15 (+) infection by driving Tox1-Snn1 ETS of the host through elevated Tox1 expression. Moreover, in two Snn1 wheat lines (H086 and H336) the total biomass at four dpi of co-infection was slightly but significantly elevated (P < 0.05) compared to SN15 (-) ( Fig 5). This suggests that SN15 (-) assists SN15 (+) in host colonisation possibly through a mutualistic interaction.

Tox1 repression by PE401 alleviates epistasis of a major SNB QTL on chromosome 2A
Tox1 is epistatic to the SNB QTL Qsnb.cur-2AS1 detected on chromosome 2A in addition to Tox3-Snn3B1 [22]. Since neither the NE or host receptor on chromosome 2A have been cloned, epistasis can only be studied through genetic mapping using wheat mapping populations that segregate for Qsnb.cur-2AS1. Therefore, interval mapping and QTL analysis were undertaken using the markers previously developed for the CxW population [22] to explore any differences in the disease interactions produced by the SN15 (-) and (+) PE401 mutant strains. The QTL detected on chromosome 1B, where Snn1 is located, explained the highest phenotypic contribution to SNB in the SN15 (-) mutant (47.9%) compared with SN15 (+) (30.1%) (Table 3). However, the Qsnb.cur-2AS1 SNB QTL was detected in the SN15 (+)

Promoter substitution identifies the positive regulatory region of Tox1
To characterise the core regulatory regions involved in gene expression, we generated SN15 promoter replacement mutants that carry sequentially truncated Tox1 promoter sequences (Fig 6). The shortest truncation included only the Tox1 5' UTR (untranslated region) from the annotated genome which encompassed 86 bp upstream of the Tox1 start codon [35]. Further truncations were made at 107 bp and 157 bp upstream of the ATG, either side of a TATATAA sequence typical of a core eukaryotic TATA box element [49]. Additional truncations were made at 263 bp and 310 bp upstream to exclude/include the site of the PE401 variant. In addition, 598 bp and 873 bp upstream mutants were selected to exclude possible undiscovered regulatory elements; the 873 bp region corresponding to the entire sequence of the SN15 (-) mutant used in the preceding analyses. Furthermore, SN15 (+) mutants carrying PE401 were generated in each of the 310, 598 and 873 bp mutant backgrounds to assess any effect of the upstream elements on the repressor activity ( Fig 6A). Gene expression analysis revealed that rather than any single element being essential, an additive effect on Tox1 gene expression was observed by the inclusion of regions up to 310 bp, where the expression was maximal (Fig 6B). In particular, significant increases were observed by the inclusion of the putative TATA box region (107-157 bp) and the upstream region (157-263 bp). PE401 acted as a strong repressor in each of the promoter deletion mutants tested. The inclusion of longer promoter regions did not significantly increase Tox1 gene expression. This suggested that positive regulatory elements are located directly adjacent to the 3' end of the PE401 which functions as a repressor sequence independent of upstream sequences.

Discussion
In this study, we sought to characterise the gene regulatory elements controlling Tox1 expression and explore their role in NE epistasis and the virulence of P. nodorum. An important finding was the clear role in Tox1 repression conferred by PE401 located 267 bp upstream of the start codon of Tox1 (Fig 6). Analysis of the truncated promoter mutants indicated PE401 could repress the transcriptional activators driving Tox1 expression (Fig 6). The exact mechanism remains obscure, as there were no distinctive features of a non-coding element or MGE. Nonetheless, we can conclude that a novel polymorphic NE regulatory element has been identified in P. nodorum, likely targeted by sequence-specific repressor proteins.
Genetic studies indicate that field populations of P. nodorum possess a high level of genetic diversity contributed from frequent gene flow that occurs between populations in most cases.

PLOS PATHOGENS
A mechanism of effector epistasis In addition, random mating, large effective population sizes and diversifying selection of virulence loci were frequently observed in P. nodorum [1,23]. Hence, it is generally considered that P. nodorum possess a high evolutionary potential that allows the fungus to adapt to

PLOS PATHOGENS
A mechanism of effector epistasis environmental changes [1]. Therefore, there have been historical interests in understanding the population structure and evolutionary history of P. nodorum summarised in [23]. In a recent study, Pereira et al.
[50] used a pan-genomics approach to perform a population structure analysis from an international collection of P. nodorum isolates. Four distinct genotype groups were observed consisting of Australia/South Africa, Europe/Southern US, Northern US/Iran and Oklahoma. Similarly, Phan et al.
[47] analysed the population structure of Australian P. nodorum using microsatellite markers and observed significant changes in pathogen genotype and aggressiveness which coincided with shifts in the mass adoption of commercially planted wheat cultivars over four decades. The most parsimonious explanation regarding the origins of PE401 is that the sequence was present in the ancestral Tox1 gene promoter and was subsequently lost in a subset of isolates. This stems from the fact that it was only detected in isolates outside the proposed centres of origin for SNB [44,45]. Regions with the highest proportion of isolates without PE401 were from South Africa and Australia. There is little data regarding South African wheat but it is known that Snn1 wheat in Australia has been widely sown throughout the period corresponding to the isolate collections assessed [24,47]. No detectable shifts in the frequency of the element were observed during this time 2001 and 2015 (S1 Fig). This indicates either that selection was neutral throughout this period, or that the maintenance of some variation in the population is beneficial for the pathogen. This situation is supported by the results of our coinfection assay, which demonstrated PE401 did not cause a fitness penalty in the presence of the strain lacking PE401. The other major P. nodorum population where PE401 was absent is the North Dakota (Northern US) population, albeit at a reduced frequency (27%) compared to Australia (87%) and the South African field sample (100%). In this region, it has been reported that Snn1 is prevalent in widely grown durum wheat varieties [16,51]. Interestingly, P. nodorum population structure analyses have suggested the North Dakota population is unrelated to isolates collected in Australia and South Africa, which are closely related to each other [44,50]. It is therefore possible that PE401 was lost on separate occasions in these populations. On the other hand, it is also possible that genetic exchange has occurred between the populations followed by selection. Both possibilities are speculative and require further investigation. Accordingly, if Snn1 is removed from cultivars sown in these and other regions, it will be important to monitor the frequency of the element in addition to the Tox1 gene in fungal isolates.
The infection assay using the CxW DH population demonstrated that higher expression of Tox1 conferred a higher average level of virulence on Snn1 wheat. While it is known that a fungal NE repertoire is associated with phytopathogenicity [13,16,18], the role of gene expression had not been extensively examined [52]. In the case of the tan spot of wheat fungus Pyrenophora tritici-repentis, a much higher level of ToxA expression than in P. nodorum is maintained during infection [53]. It was proposed this can compensate for the reduced potency of the PtrToxA isoform relative to P. nodorum variants [53] and demonstrates an example in another pathosystem where NE expression change can have significant disease outcomes.
The genetic mapping results comparing SN15 (-) and SN15 (+) on the CxW DH wheat population suggested removal of PE401 did not affect Tox1-Snn1 epistasis over Tox3-Snn3B1. This indicated that Tox1-Snn1 must entirely be removed before Tox3-Snn3B1 compensates for virulence. However, the Qsnb.cur-2AS1 SNB QTL was only detected by the inclusion of PE401. The physical interval for Qsnb.cur-2AS1 is notably large [22]. Nevertheless, several recent studies suggest the locus contains a gene conferring sensitivity to an undiscovered NE, that functions as an important determinant of both seedling, adult leaf and glume SNB resistance since these QTL overlap with Qsnb.cur-2AS1 [20,22,26,54]. In addition, a minor seedling resistance QTL on chromosome 2A was detected in a recombinant inbred population derived

PLOS PATHOGENS
A mechanism of effector epistasis from Swiss winter wheat varieties Arina and Forno [20]. Therefore, the Tox2A-Qsnb.cur-2AS1 interaction provides a layer of NE redundancy that maintains virulence on wheat, albeit at a lower level, in the absence of other interactions such as Tox1-Snn1. In addition to SNB, a colocalised Qsnb.cur-2AS1 QTL is a major contributor to tan spot resistance in wheat [55]. Interestingly, tan spot is the most damaging necrotrophic fungal disease of wheat in Australia [56] and the fungus frequently co-infects wheat with P. nodorum [57-59].
Regional P. nodorum populations differ greatly in NE gene frequency [60,61], possess high haplotype variability [45], differential isoform activity [52], variable NE expression [62; this study] and high evolutionary potential [1]. This makes SNB a difficult disease to eliminate using a common genetic resistance background in wheat across different cereal growing regions where SNB is endemic. PE401 provides a novel example of a population variant regulating NE epistasis through gene repression. Epistasis between NE-host gene interactions was frequently observed through the use of association mapping with wheat genetic mapping populations. A complex picture has now emerged as some interactions are additive in SNB [23] while others are epistatic (Table 1). Several instances of epistasis, where one fungal NE-host receptor interaction suppresses the disease contribution of another, are now documented (Table 1). In some cases, differential NE gene expression underpins epistasis, but the regulatory elements were unclear. In one example, the expression of ToxA was reported to be two-fold higher during infection in the P. nodorum isolate SN5 compared with SN4 [62]. Higher ToxA expression correlated with a greater contribution of ToxA-Tsn1 to the disease phenotype of SN5 on the BR34 x Grandin RI wheat population that segregates for Tsn1 and Snn2. Conversely, the phenotypic contribution of Tox2-Snn2 was reduced [62]. Recently, a gene encoding the NE Tox267, which interacts with Snn2, has been cloned and functionally characterised [17]. The deletion of Tox267 in SN4 resulted in up-regulation of the NEs ToxA, Tox1, and Tox3, demonstrating further connectivity between the expression of these genes. Another study that used an RI wheat population (ITMI) segregating for Tsn1, Snn1 and Snn3 reported that the presence of ToxA in SN2000 suppressed Tox1-Snn1 ETS [31]. Gene deletion of ToxA produced only a mild reduction in disease severity, which was compensated by increased Tox1 expression. Collectively, these studies suggest gene repression as the basis for NE epistasis is a common mechanism, for which the results presented here have provided a novel mechanistic insight.
The phenomenon of NE epistasis is also observed in other fungal-plant pathosystems, dictating the contribution of effector-susceptibility/resistance gene interactions during disease [27]. P. tritici-repentis possesses a near-identical copy of ToxA to P. nodorum and infects the host through Tsn1-mediated ETS [13]. The deletion of ToxA increased the virulence of P. tritici-repentis on some Tsn1 wheat lines and unmasked the effect of a novel chlorosis-inducing factor [63,64]. In the Brassicae black leg Leptosphaeria maculans-Brassica napus pathosystem, Rlm3-mediated resistance through recognition of the avirulence effector protein AvrLm3, was suppressed by another avirulence effector AvrLm4-7 [65].
Further evidence was provided here using a competitive De Wit replacement assay that demonstrated a direct selective advantage in fungal proliferation in the absence of the PE401 on Snn1 wheat lines since the high-expressing Tox1 variant confers a virulence advantage over low Tox1 expression. In Australia, Snn1 is continually being maintained in wheat varieties grown in Western Australia (WA), a region where SNB is endemic [47]. Association mapping analyses of wheat infected with P. nodorum from Australian field trials frequently detect a QTL on chromosome 1B where Snn1 is located [22, [66][67][68][69] summarised in [23]. This suggests that the Tox1-Snn1 interaction plays a role in SNB in Australia. Therefore, it is paramount to remove Snn1 from commercial wheat to improve their genetic potential for SNB resistance. Maintenance of high Tox1 expression can be bioenergetically taxing if it does not confer a

PLOS PATHOGENS
A mechanism of effector epistasis selective advantage for the pathogen. Therefore, it remains to be seen if the complete removal of Snn1 in commercial wheat through germplasm improvements will affect the frequency of PE401 within the Australian P. nodorum population in the future. Tox1 and the other two well-characterised NEs ToxA and Tox3 are present in almost all P. nodorum isolates in Australia but in other parts of the world, ToxA, 1 and 3 frequencies are much lower [45,60]. Outside of Australia, recent association studies using both GWAS and biparental mapping detected the 1B SNB QTL from field trials conducted in Norway [25,70], but the detection of Tox1-Snn1 interactions in field studies can be variable [summarised in 23]. For instance, it was previously observed that an eight-way Multiparent Advanced Generation Inter-Cross (MAGIC) population segregated for Tox1 sensitivity at the 1B locus where Snn1 is situated [71,72]. However, Lin et al.
[54] carried out a field study using a sub-set of MAGIC lines naturally infected with P. nodorum stubble across Norway and the UK during a four year trial period. Surprisingly, the Tox1-Snn1 interaction was not detected in SNB of wheat across all the trial periods. Interestingly, QTLs located on chromosome 2A were detected across all trial seasons [54]. There are several possible explanations. Firstly, natural isolates present at both trial sites may lack Tox1 [60]. A recent survey of the Norwegian P. nodorum population revealed that Tox1 prevalence in isolates collected from different sampling sites varied from 7% to 45% [61]. Second, other NE-Snn interactions may be active and epistatic to Tox1-Snn1 (Table 1). Lastly, isolates at both trial sites may possess PE401 and thus represses the role of Tox1-Snn1 in SNB. Hence, we can only speculate that there is less dependency for Tox1-Snn1 ETS outside of Australia. It was also hypothesised by Haugrud et al.
[30] that P. nodorum down-regulates Tox1 as a means to prevent antagonism with the activity of other NEs that subvert alternative host-immune responses. This hypothesis requires further testing.
From the co-infection assay, we anticipated that SN15 (-) will outcompete SN15 (+) when infecting Snn1 wheats when all other known NE-host dominant susceptibility gene interactions are removed in CxW DH lines. In contrast, SN15 (-) appears to 'rescue' SN15(+) when co-infecting. In two of the three Snn1 wheat lines, the pathogen complex accumulated more biomass than a single infection. McDonald et al. [73] observed different genotypes of P. nodorum on a single infected wheat plant which suggests that P. nodorum exists as a fungal disease complex to ensure maximal infection capabilities on susceptible wheat. This observation accounts for a mixed Tox1 expressing genotype in the Australian P. nodorum population where the high expressing variant remains dominant [47 and S1 Fig].
In this study, we observed that Tox1 exhibits both 'selfish' and 'selfless' (altruistic) characteristics. We hypothesise that the loss of PE401 allow the gene to exert dominance or epistasis over both Tox3-Snn3 and Tox2A-Qsnb.cur-2AS1 interactions which is driven by Snn1 and results in Tox1 exhibiting a 'selfish' character [74]. However, our coinfection competition assay demonstrates evidence of altruism at the isolate level where one isolate must have a higher Tox1 expressing variant to help maintain the virulence and prevalence of isolates with lower Tox1 expression, which is mediated by the presence of PE401 in the P. nodorum population. Hence, this study has provided a unique mechanistic insight into a complex NE-NE interaction that is occurring within the P. nodorum population.
The identity of trans-acting regulatory molecules that are associated with Tox1 regulation remains unknown at this stage. However, key results from this study allow us to propose a hypothetical regulatory model for Tox1 and NE-Snn epistasis detected using the CxW population (Fig 7). In the absence of the PE401, transcriptional activator(s) can drive high Tox1 expression (Fig 7A). As a result, Tox1-Snn1 contributes strongly to SNB while the Qsnb.cur-2AS1 QTL is suppressed. An unidentified repressor binds to the PE401 and inhibits the transcriptional activator(s) bound to the TATA-region [75,76], thereby repressing Tox1 expression and the consequent epistatic effect on the Tox2A-Qsnb.cur-2AS1 interaction during SNB ( Fig 7B). Although the proposed hypothetical regulatory mechanism remains speculative at this stage, it provides a framework that will allow future studies of regulatory components. As such, affinity purification assays using these sequences as bait represent a promising avenue to identify the regulators involved.

Conclusions and implications
The findings presented here have advanced our knowledge regarding the control of Tox1 expression in P. nodorum and its profound epistatic effect on SNB resistance in wheat. Although the proposed regulatory mechanism remains speculative at this stage, we have defined individual regulatory elements within the promoter region of Tox1. This provides a framework for future studies into the components regulating Tox1. We are currently using a DNA-protein bioassay approach to identify transcriptional activator and repressor proteins that are associated with the promoter region of Tox1 and regulate its expression. However, underlying questions on the mechanism of the epistatic suppression of NE gene expression and phenotypic contributions of SNB remained unanswered. Small interfering RNA may be present within the fungus that regulates gene expression including silencing NE genes [77]. Alternatively, NE-host gene interactions within the same system compete for common plantsignaling pathways, such as the mitogen-activated protein kinase pathway, or are antagonistic in the mechanisms by which ETS is triggered [15,78].
The evidence presented highlights the general importance of understanding NE regulation, the outcomes of which extend beyond the individual NE-host receptor interactions. The evolution of changes in NE gene expression provides a mechanism for P. nodorum to adapt where the complement of host receptors shifts through resistance breeding efforts. More importantly, the identification of PE401 will help to formulate effective crop protection strategies against SNB of wheat. Downie et al.
[23] provided a framework of management strategies to minimise the impact of SNB. This includes establishing an extensive isolate collection programme, epidemic monitoring, population genotyping and wheat variety adoption strategies. Based on the outcome of this study, we propose additional guidelines for SNB management to complement the Downie et al.

PLOS PATHOGENS
A mechanism of effector epistasis 1. Monitor for changes in the frequency of PE401 in regional P. nodorum populations.
2. The maintenance of Snn1 in wheat is a likely driver for P. nodorum to retain PE401 and Tox1 evolution. Therefore, Snn1 removal should be prioritised in cereal breeding programmes through conventional breeding, genome editing or genetic modifications.
3. Introduce the resistance allele for Qsnb.cur-2AS1 with other SNB resistance alleles (eg. snn1 conferring Tox1 insensitivity) into wheat breeding stocks. Phan et al. [9] and Lin et al. [54] demonstrated that stacking of four regionally relevant SNB resistance alleles provide optimal disease resistance in wheat.

4.
Qsnb.cur-2AS1 also confers resistance to tan spot [55]. Inclusion of the Qsnb.cur-2AS1 insensitivity/SNB resistance allele in breeding programmes should reduce the impact of the SNB-tan spot disease complex.
We anticipate that the outcome of this study will drive a greater level of research into the field of effector regulation and epistasis in other fungal-plant pathosystems to generate similar outcomes to improve existing crop protection strategies.

Fungal culture and isolate sources
All P. nodorum strains were maintained on V8-PDA agar at 21˚C under a 12-h photoperiod or grown in Fries3 broth as previously described [53,79]. All isolates analysed in this study were obtained from random, transects or hierarchical sampling of natural field infections as previously described [47,50,80,81]. Additional information on the sampling locations, dates collected and other relevant information available for the respective isolates is summarised in S1 File.

Compilation of Tox1 promoter variants
The P. nodorum Tox1 sequence was derived from the genome annotation SNOG_20078 of the Australian reference isolate SN15 [35]. A BLAST database was built (Geneious Prime 20.2.5) from the other available published and assembled genomes for 22 P. nodorum isolates [36,37]. The corresponding Tox1 loci, including the 1,500 bp upstream regions, were retrieved using the SN15 Tox1 nucleotide sequence. The 20 sequences retrieved were aligned using the Geneious aligner on the default settings to detect polymorphisms in the promoter region. A collection of 159 Australian isolates from a previous study [47] were then PCR screened (Tox1_screen_F/R) to detect product size shifts corresponding to a 401 bp element presence/ absence variant. All primer sequences are shown in S1 Table. Sanger sequencing was undertaken on the region for two Australian isolates with the 401 bp element (WAC13443 and WAC13072) and an additional isolate without (WAC13690). The Tox1 locus and 1,500 bp upstream regions were also retrieved from genomes that were previously assembled for a worldwide collection of 146 P. nodorum isolates (which included 20 additional Australian isolates) [81,82]. The NCBI BLAST service [83] was then used to query the sequence read archives for 184 Illumina sequenced genomes for USA isolates deposited as PRJNA398070 [80], which were non-redundant to the isolates already compiled. The Tox1 nucleotide sequence and 1,000 bp promoter from SN4 was used to query the respective archives for each isolate. Where continuous coverage of reads >95% identity was obtained across the corresponding region, the presence/absence of PE401 was able to be determined. Isolate details and the data sources used for promoter sequencing and alignments are summarised in S1 File.

PLOS PATHOGENS
A mechanism of effector epistasis

In silico analyses of the Tox1 promoter variants
A world distribution map was produced from the geographical information assembled for a panel of 484 P. nodorum isolates with or without the Tox1 PE401 (S1 File) using the maps and ggplot2 packages in R [84]. The SN4 Tox1 promoter sequence was then queried for matches in Repbase by submitting the 1500 bp region through the CENSOR portal using the default setting [85,86]. The 401 bp SN4 Tox1 PE401 was also submitted to the Rfam portal [87]. Furthermore, the NCBI BLAST service [83] was used to scan the PE401 sequence against the SN15 genome, the NCBI nucleotide and the whole-genome shotgun collections. Dot plots and GC% graphs were produced using Geneious. The SNOG_30065 expression and gene annotation data were obtained from a previous study [42].

Generating Tox1 promoter replacement mutants
The cloning strategy and an overview of the modified Tox1 promoter locus in fungal transformants are depicted in S3 Fig and the primer details are listed in S1 Table. In summary, promoter replacement constructs were assembled in two stages by attaching flanking regions to a pTef1-HygR-tTef1 resistance marker using a Golden Gate' (GG) style cloning system [88] established in-house using a pUC19 vector backbone (New England Biolabs, Ipswich, Massachusetts, USA) and kindly provided by Dr. Jordi Muria-Gonzales. In stage one the pTef1-HygR-tTef1 marker was assembled by BbsI-mediated cloning using Tef1 promoter and terminator regions derived from Aspergillus nidulans using pTef1_P_BbsI_F/R and tTef1_T_BbsI_F/R, as well as the hygromycin resistance gene amplified from Pan7 [89] with HygR_B_BbsI_F/R. Stage two involved the assembly and attachment of distinct flanks to the marker cassette using BsaI cloning. The left flank was the same for all constructs, amplified from SN15 using pTox1_HR_FL_Bsa1_F/R. The right flanks were amplified from SN15 gDNA for the (-) PE401 construct variant using pTox1_873_HR_FR_Bsa1_F/ pTox1_HR_FR_Bsa1_R. The same primers were used in combination with pTox1_Dom_B-saI_F/R to amplify the right flank from WAC13443 gDNA for PE401 constructs (which allowed fragment domestication for GG cloning by the introduction of a point mutation). PE401 was also substituted in the previous construct by amplifying the GG plasmid with pTox1_indel_3_BbsI_F/pTox1_indel_5_BbsI_R and ligating in a 401 bp spacer (S) element with a similar GC content but an unrelated sequence (S2 File). This was amplified from the pGEM-T-Easy vector (Promega, Madison, WI) using the primer pair spacer_BbsI_F/R. The three resulting homologous recombination (HR) constructs, were amplified and purified from these templates using pTox1_HR_FL_F/R, before polyethylene glycol (PEG) mediated transformation into SN15 and WAC13443. Truncated versions of the Tox1 promoter replacement construct were produced by modifying the forward primer used to amplify the right flank to incorporate either 86, 107, 157, 263, 310 (+/-PE401) or 598 bp (+/-PE401) regions upstream of the translational start site. The resulting constructs were PEG-transformed into SN15 as previously described [90]. All cloned constructs were verified through PCR screening and Sanger sequencing, while the fungal transformants were PCR screened using pTox1_HR_screen_F/ tTox1_HR_screen_R and by qPCR (Tox1_qPCR_F/R vs Actin_qPCR_F/R) to confirm singlecopy integration using a robust quantitative PCR method [91]. Two single copy mutants for all transformants were retained for gene expression analyses. All genetically modified strains and their background strains are described in Table 4.

Gene expression analysis
P. nodorum cDNA was synthesised from the respective strains/mutants grown for three days in sterile Fries3 liquid medium for three days shaken at 120 rpm in the dark as previously

PLOS PATHOGENS
A mechanism of effector epistasis described [53]. Quantitative RT-PCR was used to determine the expression level of Tox1 normalised to Act1 (2 dCt ). For in planta time series analysis of gene expression, lesions were harvested at 3, 5 and 7 days post-infection by the seedling spray method, from which cDNA was synthesised for qPCR as previously described [92]. Tox1_qPCR_F/R and Tox3_qPCR_F/R were used to assess the respective gene expression normalised to Act1 (2 dCt ). A one-way ANOVA with Tukey-HSD post-hoc test was used to test for differences (p<0.05) between isolates and/or mutants (SPSS version 27.0).

Association mapping and QTL analysis
The CxW DH wheat population that consisted of 177 lines (Intergrain Pty Ltd, Perth, Australia) was used to carry out association mapping and QTL studies of seedling SNB as previously described [22]. A total of 482 DarT, wheat genomic SSR and EST-SSR markers were used to construct a genetic map for the population [22]. Two-week old seedlings were infected with P. nodorum using a whole plant spray assay was visually assessed for SNB severity on a standardised scale of one to nine as previously described [22]. A score of one indicates no disease symptoms whereas a score of nine indicates a fully necrotised plant [9,22]. QTL mapping was undertaken using MultiQTL v. 2.6-Complete software (MultiQTL Ltd, Institute of Evolution, Haifa University, Israel). The QTL mapping was based on a genetic linkage map previously built for the CxW population using the Kosambi mapping function using MultiPoint v. 3.2 (MultiQTL Ltd, Haifa University, Israel) from maximum recombination frequencies of 0.35 [22]. This included 385 markers polymorphic between the CxW parent lines. Seedling disease scores taken from the average of three randomised biological replicates were used for interval mapping to determine QTL linked to SNB. Markers with the logarithm of the odds (LOD) score set at a > 2.5 cut-off was used to construct an interval model for the corresponding QTL as previously described [22].

PLOS PATHOGENS
A mechanism of effector epistasis

Attached leaf infection assay
An attached leaf assay with modifications was used to carry out multi-strain co-infection [93]. Briefly, the first leaves of two-week old wheat seedlings were inoculated with a P. nodorum spore suspension [1 x 10 6 spores/ml in 0.02% (v/v) Tween 20] using an airbrush kit (6ml/min, Ozito, Perth, Australia). Per treatment (+PE401; -PE401; 50:50 mix) a total of 6 ml spore inoculum was evenly sprayed across five wheat lines with 6 seedlings per line. Tween was used as a negative control. The wheat lines were H213, H324 (snn1, snn3b1, insensitive to Tox2A) and H086, H336, H091 (Snn1, snn3b1, insensitive to Tox2A). Leaves were allowed to dry and incubated in a controlled growth chamber as described previously [93]. Two and four days postinoculation, infected leaves were harvested with a scalpel and snap-frozen in liquid nitrogen (three biological repeats per time point per wheat line per treatment). DNA was extracted using phenol/chloroform method as described previously [94] and concentrations were measured using a Qubit Fluorometer (ThermoFisher Scientific, MA USA). All samples were diluted to a concentration of 10 ng/ul DNA.

Droplet digital PCR (ddPCR) quantification for fungal biomass
DdPCR was used for absolute quantification of biomass base on fungal DNA copy abundance. This approach allows for the precise and absolute quantification of DNA biomass as it measures and analyses about 20,000 reactions (droplets) in a single well [95]. We used two primer sets, a specific SN15 (+) primer targeting PE401 (pTox1_401_qPCR_F and pTox1_401_qPCR_R) and a general SN15 primer (ddPCRf and ddPCRr) that will detect both SN15 (-) and SN15 (+) DNA (S1 Table). In a single well, all the positive droplets carrying a copy of the target sequence are counted and expressed as the number of copies per input DNA concentration (in our case, 50 ng of total DNA). As such, we determined and compared the number of copies between the treatments and wheat lines as an accurate measure of pathogen biomass. We used the Evagreen (BioRad, IL USA) approach as described previously [96]. Briefly, a 96-well plate was loaded into the QX200 Auto DG to generate droplets (Bio-Rad, IL USA) in each well. Next, plates were sealed using the PX1 PCR Plate Sealer before continuing to PCR (C1000 Touch Thermal Cycler, Bio-Rad, IL USA) which was performed with the primer sets described above using the following program: 95˚C for 5 min, 44 cycles of 95˚C for 30 s, 63˚C for 30 s, 72˚C for 30 s and finish with 4˚C for 5 min followed by 90˚C for 5 min.
The plate was subsequently loaded into the QX200 Plate Reader (Bio-Rad, IL USA) and data was analysed using the QuantaSoft Software (Bio-Rad, IL USA). Genomic DNA of SN15 (-) and SN15 (+) strains were included as positive controls and additionally, primer efficiency was tested, thereby ensuring comparability between wells. A one-way ANOVA with posthoc Tukey-HSD testing was used to identify significant (P < 0.05) differences between the three infection treatments (

PLOS PATHOGENS
A mechanism of effector epistasis