Genomic tillage and the harvest of fungal phytopathogens

Authors

  • Richard Oliver

    Corresponding author
    • Australian Centre for Necrotrophic Fungal Pathogens, Department of Environment and Agriculture, Curtin University, Bentley, WA, Australia
    Search for more papers by this author

Author for correspondence:

Richard Oliver

Tel: +61 8 9362 7872

Email: Richard.oliver@curtin.edu.au

Abstract

Contents

 Summary1015
I.Introduction1015
II. Magnaporthe oryzae 1016
III. Stagonospora nodorum 1017
IV. Fusarium graminearum 1018
V.Fusarium solani, Fusarium oxysporum and Fusarium verticilioides1018
VI. Mycosphaerella graminicola 1019
VII. Leptosphaeria maculans 1019
VIII. Blumeria graminis 1019
IX.Genomic tillage and the crop of effectors1020
X.Research priorities1020
 Acknowledgements1021
 References1021

Summary

Genome sequencing has been carried out on a small selection of major fungal ascomycete pathogens. These studies show that simple models whereby pathogens evolved from phylogenetically related saprobes by the acquisition or modification of a small number of key genes cannot be sustained.The genomes show that pathogens cannot be divided into three clearly delineated classes (biotrophs, hemibiotrophs and necrotrophs) but rather into a complex matrix of categories each with subtly different properties. It is clear that the evolution of pathogenicity is ancient, rapid and ongoing. Fungal pathogens have undergone substantial genomic rearrangements that can be appropriately described as ‘genomic tillage’. Genomic tillage underpins the evolution and expression of large families of genes – known as effectors – that manipulate and exploit metabolic and defence processes of plants so as to allow the proliferation of pathogens.

I. Introduction

Fungal pathogens of crop plants remain major problems in all areas of the world, even the most developed. The study of crop disease losses has gone out of fashion, but recent studies in Australia have shown that disease losses in wheat, barley and legumes currently average c. 2–5%, 5–10% and 10–50%, respectively (Murray & Brennan, 2009, 2010, unpublished). Australian agriculture is mainly of low intensity and operates in a generally dry climate with low disease pressures, so we can expect the numbers to be even higher elsewhere. Such crop losses are a major factor limiting food security (Gurr & Rushton, 2005).

Research aimed at limiting the impact of fungal diseases has had a strong element of pragmatism. Various cultural practices (rotations, tillage, sowing timing) have focused on minimizing the exposure of the plant to the pathogen. Genetic disease resistance is exploited by the selection of resistant germplasm and attempts to combine resistance with other desirable phenotypes. Chemical fungicides mainly target core functions common to all fungi – respiration and growth.

These pragmatic approaches have at best kept up with an evolving plethora of fungal pathogens. Historical studies of phytopathology have chronicled the apparently inevitable cycle of new crop species, locations, cultivars and fungicides followed by new species or pathotypes causing novel diseases (Large, 1940; Money, 2006). Many new pathogen species have emerged in the last century or so (Oliver & Solomon, 2008) and these pathogens threaten not only crops but also animals and even entire ecosystems (Fisher et al., 2012).

The application of molecular techniques to plant pathology promised, according to the faithful cognoscenti, to end this recalcitrant picture. Molecular plant pathology can be said to have been initiated by the discovery of the first pathogen effector genes – then called avirulence factors (Staskawicz et al., 1984). The first fungal effector genes were isolated shortly afterwards (Walton, 1987). The implicit hypothesis of much work in this era was that phytopathogens differed from their sibling saprobic species by virtue of the acquisition of a handful of key genes that conferred pathogenicity. Pathogenicity was assumed to be an advanced character, derived from saprobic predecessors. The acquisition of these key genes was assumed to involve conventional processes of vertical evolution; that is, the minor modification of the structure and expression of the genes such that they enabled growth within a plant. The field was puzzled by the demonstrable fact that many pathogens expressed genes that, by themselves, caused normally pathogenic species to become nonpathogens – the avirulence genes. Yet further confusion stemmed from the accepted division of fungal pathogens into two classes: biotrophs, which expressed the paradoxical avirulence genes; and necrotrophs, which appeared to express a nonspecific cocktail of toxic and biopolymer degradation gene products. Furthermore, there was a third class that mysteriously combined the features of the first two: the hemibiotrophs. Why should pathogens express genes that prevent pathogenicity; if necrotrophs merely expressed a nonspecific cocktail of weapons, why were they host-specific and even cultivar-specific; and how could these two apparently incompatible patterns be combined successfully in the hemibiotrophs?

The answers to these questions began to emerge during the last 25 yr in response to efforts based on gene-by-gene functional approaches. Genome sequencing of fungi began in 1996 with the Saccharomyces cerevisiae genome (Goffeau et al., 1996). The first filamentous fungal genomes were the model species Neurospora crassa (Galagan et al., 2003) and Aspergillus nidulans (Galagan et al., 2005) and the pathogen Magnaporthe oryzae (Dean et al., 2005). Since then, c. 100 pathogen and nonpathogen species have been sequenced, although only a small fraction have been published to date (Hane et al., 2011). The purpose of this review is to discuss how the study of these genomes has illuminated our knowledge of pathogenicity. The picture that emerges is one of substantial genomic rearrangements, chromosome duplications, horizontal gene acquisition and wholesale gene evolution caused by transposon proliferation and repeat-induced point (RIP) mutations. The scale of the genome alterations is such that the term ‘genomic tillage’ seems particularly appropriate. In the field, tillage is the practice of ploughing and thorough mixing of soil before planting. It appears that the genomes of these pathogens have been shuffled and rearranged as thoroughly as any seed bed. I also note that the pathogens that are subject to genomic tillage seem to be favoured by zero- or limited tillage practices (Oliver & Solomon, 2008). The genomes of these species are discussed in the following.

The processes of genomic tillage are not restricted to fungal plant pathogens. All organisms are subject to duplications, horizontal gene transfer and transposition (though only fungi have RIP) to greater or lesser extents. Genomic tillage processes generate a large pool of highly modified mutant progeny populations that can be selected by evolutionary processes. Changes in agriculture have accelerated over the last 150 yr and these changes have created novel niches into which mutant pathogen populations can expand. Rapid environmental change favours those organisms that are subject to wholesale mutagenic processes by allowing the rapid proliferation of selected lines whilst the vast bulk of variants die out.

The term ‘effectors’ has been adopted to describe all the gene products that alter (positively or negatively) the interaction with the host plant. These include the previously termed avirulence and host-specific toxin genes (Oliver & Solomon, 2010; Koeck et al., 2011; Thomma et al., 2011). Genomic tillage processes are linked to the inter- and intraspecific diversity of effectors.

This review focuses attention on the wholesale nature of genomic differences between species and isolates and considers how this affects research priorities for the future. The scope is limited to ascomycete plant pathogens for reasons of space, but similar arguments can be forwarded for both basidiomycete and oomycete groups. Two other reviews have appeared during the preparation of this article. They both highlight the unexpectedly dramatic amount of genome structural innovation in fungal plant pathogens (Raffaele & Kamoun, 2012; Spanu, 2012). Both also cover the Basidiomycete genomes.

II.  Magnaporthe oryzae

The first draft of the genome sequence of M. oryzae (syn. grisea) strain 70-15 was obtained in 2002 and was published in 2005 (Dean et al., 2005; Xu et al., 2006). This achievement represented a paradigm shift for fungal phytopathology because it was expected to give easy access to all the genes responsible for the pathogenic properties of the organism. The expectation was that the genome sequence would be completed so that the contigs represented entire chromosomes. Simultaneously, the genome sequence would be annotated with greater and greater fidelity until a complete and accurate list of protein coding genes would be available. The process of gene annotation turned out to be much more complex, and hence inaccurate, than had been predicted. This was especially true of small and secreted proteins, many of which are key determinants of pathogenicity. Expressed sequence tag (EST) sequencing using Sanger methodologies (Donofrio et al., 2006; Soderlund et al., 2006; Oh et al., 2008; Numa et al., 2009) and, more recently, entire mRNA sequencing and proteomics have added depth to the data (Wang et al., 2011).

Even before the sequence was first published, it became clear that different strains of M. oryzae differed markedly in gene content (Xu et al., 2006). Thus the stakes were raised; to understand pathogenicity and combat the losses caused by the pathogen, it would be necessary to sequence at least a sample of the M. oryzae population. This also revealed the ultimate futility of attempting to perfectly define the 70-15 genome assembly and its associated genes.

A further surprise was the lack of obvious genome-level similarity to N. crassa. These species are closely related by fungal standards, both being in the Sordariomycetes, but there was little or no chromosomal collinearity between the species. This contrasted with the situation in plants and animals where related species shared extensive collinearity at the whole chromosomal level (Moore et al., 1995). This was a disappointment because it dashed hopes of scanning the two genomes and thereby identifying missing genes that would be good candidates for key pathogenicity determinants. Instead, analysis of the genome revealed three characteristics that would become themes as further genomes were discovered. The first was the large number of orphan genes without obvious homology in other species – this number has only slowly declined as further genomes were sequenced, suggesting very rapid evolution of large numbers of gene sequences. The second was data suggesting that horizontal gene transfer has played a role in amending the gene complement of Ascomycete fungi (Richards et al., 2006; Marcet-Houben & Gabaldon, 2010; Richards, 2011). And the third was the presence of a complex array of repetitive elements and transposons.

What did we learn from the M. oryzae genome sequence? It was clear that the number of genes was more or less 12 000. This was similar to the numbers predicted for the nonpathogens Aspergillus and Neurospora and clearly significantly more than the 6000 or so found in various yeast species. One expectation was that, by comparing these three filamentous species, it would be possible to identify pathogenicity factors directly; however, all three species had a common core and a similar number of unique genes. This expectation was naïve for at least two reasons. First, the concept that pathogenicity is an ‘advanced’ group derived from ‘primitive’ saprobic species was, on reflection, unsupportable. Saprobic species mostly consume the dead remains of plants. Hence, as soon as plants evolved, both saprobes and pathogens would have emerged soon after. We must assume that the evolutionary histories of both groups are of similar longevity. Secondly, it was not clear whether either Aspergillus or Neurospora could be regarded as a representative ‘wildtype’ species (any more than Saccharomyces). Model species were chosen by the pioneer fungal geneticists for a nonrandom set of reasons: ease of growth on simple media; speed of growth; and stability of phenotype. These properties were later explained, at least for N. Crassa, by the exceptional lack of transposons in its genome, a feature explained by the efficient operation of RIP (see later for further discussion) (Galagan & Selker, 2004). Transposons had been discovered in pathogenic filamentous fungi some years earlier (McHale et al., 1992). They were common in M. oryzae (Farman et al., 1996), differed between populations from different hosts (Hamer et al., 1989) and were already known to have direct effects on virulence (Kang et al., 2001). Clearly the distinction between saprobe and pathogen was complex and individual and would need many more genomes before the distinction could be clarified.

In addition to defining new questions about pathogenicity, the M. oryzae genome also focused attention on the distinction between biotrophs and necrotrophs. The name of the disease ‘blast’ and its appearance as necrotic lesions originally placed the pathogen in the necrotroph class. However, the presence of negatively acting effectors (previously called avirulence genes) indicated that the pathogen had properties in common with rusts and powdery mildew, the undisputed biotrophs (Valent et al., 1991). Hence the pathogen was rather uncomfortably placed in the compromise hemibiotroph class (Oliver & Ipcho, 2004). Understanding the differences between these (at least three) classes would need more genomes.

The major impact of the genome sequence was that it facilitated the study of pathogenicity in M. oryzae itself. The genome sequence greatly aids the design and testing of vectors for gene ablation and allows gene expression analysis to be carried out efficiently. As a result, the number of genes studied in depth in M. oryzae continues to grow apace (Jeon et al., 2007). These studies have been reviewed extensively (Wilson & Talbot, 2009; Xu et al., 2011).

The early realization that strains differed in gene content argued for the analysis of multiple strains. Next-generation sequencing is beginning to give us a technical solution, but we already have a taste of the future from an analysis of selected genes from 46 distinct strains (Mach, 2009; Yoshida et al., 2009). Analysis of a set of putative effectors genes for ones showing polymorphisms associated with virulence allowed the identification of three novel effectors. In the future, as new sequencing technologies mature, this type of study can be expected to involve the sequencing of large numbers of new strains.

III.  Stagonospora nodorum

The genome sequence of S. nodorum (also known as Phaeosphaeria nodorum) was completed in 2005 and published in 2007 (Hane et al., 2007; Oliver et al., 2012). It was the first Dothideomycete and first true necrotroph to be published. Analysis of the gene content was aided both by a microarray experiment and by proteomic analyses (Ipcho et al., 2012) and resolved to the expected c. 12 000. A large number (c. 1300) of the genes were found to be small, and/or secreted, and/or expressed early in infection, but to have no obvious homology to any known gene. They were therefore described as effector candidates, further confirming that pathogenicity seems to require a very large arsenal of genes, rather than a handful of key pathogenicity factors.

The repetitive element content of S. nodorum was modest in size but contained no obviously intact transposon genes. This was a surprise as active transposons had been discovered in this species (Rawson, 2000). The explanation was that the transposons had been inactivated in the sequenced (Australian) strain by the process of RIP (Hane & Oliver, 2008, 2010). RIP is a fungal -specific process whereby duplicated genes are both mutated during the meiotic cycle (Galagan & Selker, 2004). Technically, proof of RIP requires analysis of the products of a cross, but the hallmarks of RIP are unmistakable and abundant. The biogeography of the species (Stukenbrock et al., 2006) indicates a migration from the Fertile Crescent radiating in historical times to the current centres of wheat production but with little exchange between current localities. The process of RIP would therefore not only destroy intact transposons in an individual but ultimately in an entire population. This would limit the evolutionary potential of the mutagenic effects of both the transposons and the processes of RIP that depend on them. As with M. oryzae, the sequencing of one strain suggested questions that can only be answered by sequencing more strains.

The major impact of the S. nodorum genome sequence was that it enabled the rationalization of nascent concept that this pathogen produced necrotrophic effectors (then described as host-specific toxins). Tox1, an isolate- and cultivar-specific effector, had been found in the culture filtrate of S. nodorum (Liu et al., 2004) but had not been purified; nor had its gene been identified. Tox1 was known to be a small secreted protein. The only other pathogen known to produce such proteinaceous effectors was the related Dothideomycete Pyrenophora tritici-repentis, which produced at least two, ToxA and ToxB. It was completely unexpected that a gene sequence 98% identical to ToxA was found in the S. nodorum genome. It was subsequently shown that S. nodorum produces several necrotrophic effectors (ToxA, Tox1 and Tox3 plus more than a dozen others), each of which makes a contribution to its virulence depending on which host sensitivity genes are present (Liu et al., 2006, 2009, 2012; Friesen et al., 2007, 2008b). These effectors have been transferred to breeders to aid in the selection of disease-resistant cultivars (Oliver, 2008).

The presence of ToxA in both S. nodorum and P. tritici-repentis leads to the hypothesis that the gene had been transferred from the former to the latter by horizontal gene transfer (Friesen et al., 2006). The startling realization was that several lines of evidence suggested that the horizontal transfer took place between 1920 and 1940 and has led directly to the emergence of what is now a major and still spreading pathogen of wheat: tan spot. This brought the horizontal gene concept out of the distant past into living memory. The mechanism is still obscure but clearly involved the transfer of at least 11kb of DNA – and probably much more.

The origin of the effector genes in S. nodorum has been studied from biogeographic data (Stukenbrock & McDonald, 2007). It is likely that ToxA (and possibly the other effectors) was not present in the species when it first evolved in the Fertile Crescent, but has also been horizontally acquired in the relatively recent past. Thus we have a picture in which S. nodorum isolates harbour various combinations of effector genes that have been acquired horizontally. These act as a source of effectors for other species with which it comes into contact.

These effectors have a positive effect on virulence when the wheat host carries the matching recognition gene. In the case of ToxA, the recognition gene, Tsn1, has been cloned and it closely resembles a biotrophic resistance gene and switch having nucleotide-binding site (NBS), leucine-rich repeat (LRR) and protein kinase domains (Faris et al., 2010). It appears that this necrotroph has sabotaged a disease or pest resistance gene by actively expressing an effector that activates cellular destruction and permits proliferation of the pathogen. Furthermore, the pathogen expresses different versions of the gene that have different levels of effector activity (Tan et al., 2012) and expresses the gene at different levels (Faris et al., 2011), presumably in order to ‘tune’ the amount of necrosis to maximize proliferation. Clearly, we can no longer sustain the concept that biotrophs are sophisticated and necrotrophs are crude (O'Connell & Panstruga, 2006; Talbot, 2010).

The effector genes of S. nodorum are all found in multiple forms and may have been subject to diversifying selection (Stukenbrock & McDonald, 2007; Liu et al., 2012). Two of the genes are found adjacent to repetitive DNA, but so far we have no evidence that RIP has been responsible for the diversification.

IV.  Fusarium graminearum

The F. graminearum genome sequence was also published in 2007 as a high-quality finished assembly and comparison with a second strain (Cuomo et al., 2007). F. graminearum is primarily found on the heads of developing wheat where its major impact is to contaminate grain with a cocktail of mycotoxins, including DON, which serve to inactivate defence responses of wheat and simultaneously render the remaining grain inedible (Desmond et al., 2008). F. graminearum could not be described either as a biotroph or as a necrotoph and so, like M. oryzae, was placed in the hemibiotroph category. Apart from the presence of the already identified DON-locus, the gene content of F. graminearum was not obviously different to either Magnaporthe or Stagonospora. Again it carried a large number of orphan genes. The major surprise emerging from the F. graminearum genome was at the whole-genome structural level. The genome was arranged in just four chromosomes and comparison of the two strains identified a clear distinction between regions high and low in polymorphisms. It seems likely that the regions of low single nucleotide polymorphism (SNP) density correspond to ancient chromosomes that have been linked end to end via the regions of high variability (Cuomo et al., 2007). It was hypothesized that the regions of high variability (described as regions of genome innovation) are a ‘workshop’ where genes required for infectivity on the host are evolved. The mechanism has remained obscure. The repetitive element was low and evidence was presented that RIP had successfully prevented the proliferation of transposons.

V. Fusarium solani, Fusarium oxysporum and Fusarium verticilioides

The Nectria haematococca genome (syn. Fusarium solani) was the second of the Fusarium supergenus and the first dicot pathogen genome to be published (Coleman et al., 2009). One prediction borne out was that the retinue of carbohydrate-active enzymes had a large number of pectate active enzymes genes and was clearly distinct from the complement of carbohydrate-active enzymes in monocot pathogens. However, it also had a much larger number of all genes than was expected: 15 707. The origins of these extra genes were obscure but apparently diverse. The presence of a large number of paralogous gene families pointed to one or more genome duplication events. The presence of c. 300 gene families not present in F. graminearum pointed to horizontal gene transfer events, a hypothesis supported by altered codon usage bias, a lower guanine-cytosine (GC) content in these families compared with the bulk genome and proximity to repetitive elements. Indeed, at least three entire chromosomes, 14, 15 and 17, can be described as supernumerary (= accessory), as they are present in only some strains. Thus entire chromosomes and sections of others have a low gene density; the genes that are carried here are nonessential but may represent a toolkit of genes and gene parts that can be used and discarded and the pathogen evolves to infect different hosts.

The F. oxysporum and Fverticilioides genomes substantiated this model (Ma et al., 2010). F. oxysporum is a complex of populations, each of which is able to infect only one of a few species of host plant. Comparative sequencing of several isolates (Rep & Kistler, 2010) showed that several entire chromosomes and smaller regions were found to be present only in a specific group able to infect one host. These were described as lineage-specific (LS) chromosomes. Such LS regions have low gene contents, with genes that tend to have obscure functions, and a high repetitive element content, suggestive of having been of horizontally acquired. Bioinformatic analyses suggest that RIP is undetectable in Fusarium spp. (Clutterbuck, 2011). Effector genes were predominantly located on the LS chromosomes and were often found to be adjacent to repetitive DNA.

Nearly all the literature pertaining to horizontal gene transfer in eukaryotes uses analyses of genomes of existing taxa to infer their evolutionary history. This study dramatically demonstrated not just horizontal gene transfer, but horizontal chromosome transfer, experimentally along with transfer of host specificity (Ma et al., 2010).

VI.  Mycosphaerella graminicola

The wheat Septoria leaf blight pathogen Mycosphaerella graminicola (Zymoseptoria graminicola) was the second Dothideomycete wheat pathogen to be sequenced and published. This species has a rich history of functional genetic (Kema et al., 2008) and biogeographic research (Stukenbrock et al., 2007, 2011). The genome sequence was published in 2011 (Goodwin et al., 2011). Just as in the Fusarium spp., there was a clear distinction between core chromosomes and accessory (supernumerary/lineage specific) chromosomes. This result had been foreshadowed using the excellent conventional genetic map available for this species (Wittenberg et al., 2009). The core chromosomes (1–13) have a high gene density, a high frequency of genes with deducible functions and a low repetitive element content. By contrast, the accessory chromosomes (14–21) have a low gene content, many of which are either of obscure function or even pseudogenes, a low GC % and a high repetitive element content. Earlier biogeographic evidence had indicated that this species had emerged when wheat was domesticated, 8000–10 000 yr ago (Stukenbrock et al., 2007, 2011). The hypothesis is that the eight small accessory chromosomes contributed to the rapid adaptation to wheat, to different wheat cultivars (Arraiano et al., 2009) and to varying climatic and cultural conditions (Zhan & McDonald, 2011) by providing the toolkit of transposons, genes and gene parts, as mentioned earlier for Fusarium.

Yet another relatively obscure mode of genomic alteration was discovered by comparison of M. graminicola with two closely related species – intron transposition (Torriani et al., 2011). The origin of spliceosomal introns in eukaryotic genes has always been a subject of intense curiosity. This study identified orthologous genes in the three species into which introns had apparently entered after the separation of these species. This study added a new dimension to our understanding of genetic polymorphism.

VII.  Leptosphaeria maculans

The oilseed rape (canola) blackleg pathogen is the major constraint to production in many parts of the world. It is a Dothideomycete fungus, closely related to S. nodorum. In contrast to the wheat pathogen, L. maculans is known to possess negatively acting effectors (avr genes) and infects canola with a prolonged latent phase (Hammond et al., 1985). In these respects, it more closely resembles M. graminicola. Three effector genes had been cloned by the time the genome sequence was released and published (Gout et al., 2006; Huang et al., 2006; Rouxel et al., 2011). The genome sequence presented yet another new and unexpected structure. The genome was found to contain 35% repetitive DNA, a significantly larger content than S. nodorum. Most of this repetitive DNA was found to comprise a complex of mostly degraded transposable elements and was markedly adenine- and thymine-rich (AT-rich). Unlike all other species considered so far, the repetitive/AT-rich DNA was collected into a series of discrete regions reminiscent of the proposed isochore structures in vertebrate genomes. Like the accessory, supernumerary and LS chromosomes of Fusaria and M. graminicola, the AT-rich regions contained few genes, many of which were of obscure functions. Remarkably, this is where some of the effector genes were found (Gout et al., 2006). Furthermore, analysis of multiple alleles of effectors in different strains revealed that the differences between them were consistent with the action of RIP (Fudal et al., 2009; Van de Wouw et al., 2010). The model that has emerged is that effector genes in successful fungal lineages evolve rapidly. This means that multiple variants are present and one or more of these may have the appropriate mutations that enable them to exploit new niches afforded by the introduction of new crops, cultivars and cultural practices. The invasion and proliferation of transposons create a wave of mutagenic activity by inserting within or nearby effector genes and promoting duplication events. A subsequent meiotic round unleashes RIP on the repetitive DNA which also affects nearby single copy effector genes. This RIP activity both inactivates genes, by introducing stop codons, and alters coding sequences by mutating Cs to Ts. This mutagenic activity promotes the evasion of defence responses by eliminating or altering effectors so as to prevent recognition by R genes.

Why has L. maculans accumulated so much more repetitive DNA than its close relative S. nodorum? Both species undergo at least one sexual cycle annually, but S. nodorum undergoes many more asexual cycles. This might suggest that rapid (about weekly) asexual cycles select against isolates with expanded genomes. However, the next pathogen indicates that this is not a sufficient explanation.

VIII.  Blumeria graminis

The obligate biotrophs represent a particular challenge for molecular biologists, as even obtaining sufficient material for genomic or biochemical analysis is a daunting task. Functional genomic tools have so far remained elusive. Nonetheless, studies of the barley powdery mildew pathogen B. graminis f. sp. hordei are amongst the most detailed of any phytopathogenic system and have created the basis for the efficient control of this disease. The genome sequence (Spanu et al., 2010) revealed a massively expanded genome of over 100 Mb, of which 64% was repetitive and transposon-derived. Despite the large genome size, the number of genes is estimated to be c. 6000, about half that of nonobligate pathogens and saprobes. Amongst the missing genes were a number of core metabolic genes involved in processes such as nitrate and sulphate assimilation, fermentation and vitamin biosynthesis. Evidently these processes are unnecessary for a strict parasite. Indeed, nitrate reduction has previously been shown not to be required for pathogenicity in a number of other pathogens (Howard et al., 1999). Also missing are genes coding for RIP processes. This evidently explains how the genome has expanded, as no mechanism exists to limit the proliferation of transposons.

The picture that emerged from the Blumeria genome was surprising on many levels. First, the metabolic load created by the replication and repair of this very large genome was tolerated and made no impact on the success of this fast-reproducing pathogen. Secondly, a large number of core genes are unnecessary if a pathogen is a strict parasite. Thirdly, and more speculatively, it is possible that massive proliferation of transposon expansion has contributed to the rapidity with which the pathogen has overcome new resistance genes.

IX. Genomic tillage and the crop of effectors

Genomic sequencing projects have contributed to a revolution in our understanding of plant–ascomycete interactions. The focus of attention has alighted on the army of effectors expressed by each pathogen. Although each pathogen has, at most, a handful of characterized effectors, their shared features have enabled us to survey the genomes for entire complement. Effectors, if defined as molecules produced by a pathogen that have an effect on the outcome of the interaction with the host, are diverse. They include both proteinaceous and nonprotein molecules. The nonprotein molecules are typically the products of polyketide synthase (PKS) or nonribosomal peptide synthase (NRPS) clusters. Genome sequences give direct access to these genes. The proteinaceous effectors share features that, by themselves, do not immediately identify the genes. They tend to be small, secreted, cysteine-rich and encoded on genes near transposons. Comparison of alleles tends to show high diversity (dN > dS). In some cases, motifs have been detected (Godfrey et al., 2010). These features can be easily used to draw up lists of effector candidates and these typically number 200–1600 (Schmidt & Panstruga, 2011). How many of these effectors will turn out to be relevant for the disease remains to be seen. Genetic evidence in S. Nodorum, for example, would suggest the number is at least 20 (Oliver et al., 2012).

The outstanding feature of effectors is their diversity between and within species. Both presence/absence and substitution polymorphisms are common (Friesen et al., 2008a; Ma & Guttman, 2008; Dodds et al., 2009; Sacristán et al., 2009). The genome sequences reviewed here show that the evolution of the effector complement involves a diverse array of evolutionary processes that have resulted in the rapid evolution of pathogenicity and allowed diseases to keep up with new agricultural practices, such as minimum tillage (Jørgensen & Olsen, 2007) with new resistance genes (Wulff et al., 2009) and may contribute to the evolution of resistance to new fungicides (Cools & Fraaije, 2008; Goodwin et al., 2011).

The evolution of effector genes after mutagenesis by conventional means such as UV radiation should not be ignored, but it seems that these simple processes are not sufficient to explain the evolution of our contemporary pathogens. Instead, we must invoke horizontal gene transfer as one of the key processes in the evolution of successful pathogens. Such horizontal gene transfer applies not only to effector genes but also to transposon sequences. Invasion of a genome by a transposon would seem an unlikely start for the evolution of the successful pathogen; and certainly we can assume the vast majority of descendants would have failed to survive. The mutagenic potential of the transposons involves at least three processes (Rep & Kistler, 2010; Raffaele & Kamoun, 2012; Spanu, 2012). First, the transposon can insert in or near a gene and thus alter its sequence and expression pattern. Secondly, the transposons can duplicate sequences as they transpose throughout the genome; this promotes variation and neofunctionalization. And thirdly, when duplicated, the transposons are subject to RIP, which not only affects the transposon but also nearby genes (Irelan et al., 1994). Thus transposon invasion allows the simultaneous creation of numerous new alleles and combinations. The population sizes of fungal plant pathogens have been shown to be large, at least in the case of M. graminicola (Stukenbrock et al., 2011), although we lack data on most pathogens. When novel agricultural practices create a monoculture of relatively clean crops, the large population sizes combined with the massive diversity is sufficient to allow the selection of a new virulent population that can exploit the new environment.

X. Research priorities

How does this picture of the genomic basis of pathogenicity affect out research directions? First, and as already mentioned, we must budget for the sequencing of not just one isolate of a pathogenic species, but sufficient individuals to capture the variation within the population. Secondly, it suggests that it is highly desirable to produce assemblies of species and isolates in which contigs are substantial fractions of entire chromosomes. The fragmented assemblies produced with short reads are useful for identifying many genes but they cannot reveal the chromosomal-scale features shown here to be critical for pathogenic evolution. Thirdly, it focuses attention on the mechanisms by which pathogens can evolve, postulated here to be the result, at least in part, of the horizontal acquisition of transposon and effector genes. How does horizontal gene transfer occur and what are its effects? Finally, it suggests we must sequence at least a sample of saprobic species that are siblings of the pathogens that have already been sequenced. The prediction is that such species might lack the cohort of effectors possessed by the pathogen.

Acknowledgements

Research in the author's laboratory is supported by the Australian Grains Research and Development Corporation. The ideas discussed in this paper have emerged in collaboration with many colleagues, who are thanked for their insight.

Ancillary