Exploitation of genomics in fungicide research: current status and future perspectives

Summary

Every year, fungicide use to control plant disease caused by pathogenic fungi increases. The global fungicide market is now worth more than £5.3 billion, second only to the herbicide market in importance. In the UK, over 5500 tonnes of fungicide were applied to crops in 2010 (The Food and Environment Research Agency, Pesticide Usage Statistics), with 95.5% of the wheat‐growing area receiving three fungicide sprays. Although dependence on fungicides to produce food securely, reliably and cheaply may be moderated in the future by further developments in crop biotechnology, modern crop protection will continue to require a diversity of solutions, including effective and safe chemical control. Therefore, investment in exploiting the increasingly available genome sequences of the most devastating fungal and oomycete phytopathogenic species should bring an array of new opportunities for chemical intervention. To date, the impact of whole genome research on the development, introduction and stewardship of fungicides has been limited, but ongoing improvements in computational analysis, molecular biology, chemical genetics, genome sequencing and transcriptomics will facilitate the development and registration of the future suite of crop protection chemicals.

Introduction

Agricultural productivity in the second half of the 20th century has continued to advance through a combination of crop genetic improvement, innovations in crop cultivation and farming practice, and advances in the effectiveness and consistency of crop protection chemicals. In the future, the demands on food production systems will be even higher. Certainly, advances in plant biotechnology, plant breeding and agronomy are central to meeting these demands. However, successfully combating pathogen‐driven diseases will require a combined approach, including the use of chemicals (Lucas, 2011).

Fungicides marketed for the control of disease on food crops are under pressure, the most predictable being the development of pathogen resistance, the consequence of a basic evolutionary process. The application of a fungicide can lead to the selection of a proportion of the population that survives, is able to resist the compound and passes this characteristic to its offspring. This phenomenon affects all fungicides, including certain protectant multisite inhibitors. Resistance to modern, systemic and highly efficacious fungicides, which have a single biochemical target, readily occurs, as a mutation in a single gene is usually sufficient to confer a highly resistant phenotype (Brent and Hollomon, 2007). Consequently, populations of some of the most destructive pathogen species are resistant to many of the currently registered antifungal chemicals (Table 1). In addition, the diversity of compounds available to growers for the control of crop diseases could be restricted in the future by European pesticide legislation, aiming to minimize the risks to human health and the environment from the use of pesticides [Sustainable Use Directive 2009/128/EC and Plant Protection Products Regulation (EC) 1107/2009]. Therefore, to be able to maintain chemical diversity, there is a pressing requirement to identify and develop new crop protection chemicals.

Table 1.

Examples of fungicide resistance reported in globally important plant pathogenic fungi and oomycetes with genomes sequenced

Pathogen	Crop	Example(s) of resistance (reference)	Genome sequence status (strain/current version/reference)	Genome centre/laboratorya
Blumeria graminis f. sp. hordei	Barley	MBCs (Vargas, 1973) DMIs (De Waard et al., 1986) Azanaphthalenes (Hollomon et al., 1997) QoIs (Heaney et al., 2000)	Partially assembled (DH14/version 1/Spanu et al., 2010)	Blumeria Genome Sequencing Consortium
Botrytis cinerea	Various, particularly grapevine	MBCs (Leroux and Clerjeau, 1985) DMIs (Elad, 1992) QoIs (Ishii et al., 2009) SDHIs (Avenot and Michailides, 2010)	Draft complete (B05.10/version 2/Amselem et al., 2011)	BROAD Institute
Fusarium spp. F. graminearum F. oxysporum f. sp. tomato F. verticilloides	Various Cereal/noncereal Tomato Cereal	MBCs (Chen et al., 2009) Phenylpyrroles (Peters et al., 2008) DMIs (Yin et al., 2009)	Assembled (Cuomo et al., 2007; Ma et al., 2010)	BROAD Institute
Magnaporthe oryzae	Rice	QoIs (Ma and Uddin, 2009) MBI‐R (Zhang et al., 2006) MBI‐D (Yamada et al., 2004)	Finished (70‐15/version 8/Dean et al., 2005)	BROAD Institute
Mycosphaerella fijiensis	Banana	MBCs (Canas‐Gutierrez et al., 2006) DMIs (Canas‐Gutierrez et al., 2009) QoIs (Sierotzki et al., 2000)	Draft complete (CIRAD86/version 2)	DOE Joint Genome Institute
Mycosphaerella graminicola	Wheat	MBCs (Griffin and Fisher, 1985) QoIs (Fraaije et al., 2005) DMIs (Cools and Fraaije, 2008) SDHIs (Fraaije et al., 2012)	Finished (IPO323/version 2/Goodwin et al., 2011)	DOE Joint Genome Institute
Phakopsora pachyrhizi	Soya bean	No reports	Draft complete	DOE Joint Genome Institute
Phytophthora infestans	Potato/tomato	PAs (Davidse et al., 1981) CAAs (Dereviagina et al., 1999)	Partially assembled (T30‐4/version 3/Haas et al., 2009)	BROAD Institute
Puccina graminis f. sp. tritici	Wheat	DMIs (Napier et al., 2000)	Finished (CRL 75‐36‐700‐3/version 1/Cantu et al., 2011; Duplessis et al., 2011)	BROAD Institute
Pyrenophora teres	Barley	QoIs (Semar et al., 2007) DMIs (Peever and Milgroom, 1992)	Draft complete (Ellwood et al., 2010)	DOE Joint Genome Institute
Pyrenophora tritici‐repentis	Wheat	QoIs and DMIs (Reimann and Deising, 2005)	Draft complete	BROAD Institute
Rhynchosporium commune	Barley	DMIs (Kendall et al., 1993) MBCs (Locke and Phillips, 1995) QoIs (FRAC, www.frac.info)	Ongoing	Wolfgang Knogge, Leibniz Institute of Plant Biochemistry, Halle, Germanyb
Stagnospora nodorum	Wheat	QoIs (Blixt et al., 2009)	Draft complete (Hane et al., 2007)	BROAD Institute

CAA, carboxylic acid amide; DMI, demethylation inhibitor; MBC, methylbenzimidazole carbamate; MBI‐D, melanin biosynthesis dehydratase inhibitor; MBI‐R, melanin biosynthesis reductase inhibitor; PA, phenylamide; QoI, quinone outside inhibitor; SDHI, succinate dehydrogenase inhibitor.

^aBROAD Institute fungal genome initiative (http://www.broadinstitute.org/scientific‐community/science/projects/fungal‐genome‐initiative/fungal‐genome‐initiative); JGI genome sequencing project list (http://genome.jgi.doe.gov/genome‐projects/); Blumeria Genome Sequencing Consortium (http://www.blugen.org/).

^bNot publicly available.

The expansion of genomic information and the advancement of technologies to exploit this information provide a novel platform from which new routes can be developed to obtain the future suite of crop protection chemicals. In 2002, after the release of the genomes of important human fungal pathogens, including Candida albicans, Aspergillus fumigatus and Cryptococcus neoformans, several genomics strategies were initially suggested to permit the rational design of antifungal compounds. These included transcript profiling, fitness tests and proteomics. However, the authors at the time predicted that these new research approaches might take a decade to bring a novel bioactive to the marketplace (Jiang et al., 2002). Since then, the genomes of many of the most important fungal and oomycete pathogens of crops have been sequenced. These are now available as advanced drafts, partial assemblies or fully assembled formats (Table 1). Typically, the genome of a selected reference strain for each pathogenic species has been assembled, annotated by the research community and is available within a genome browser and fully accessible through downloads. In addition, for some species, the genomic sequences have been aligned to the available chromosome information.

Comparative bioinformatic analysis of these sequenced fungal and oomycete genomes has already led to the identification of specific genes and/or gene families important for the lifestyle of priority pathogens and, in some species, the elucidation of the genetic landscape underpinning the evolutionary arms race with their hosts (Goodwin et al., 2011; Haas et al., 2009; Ma et al., 2010; Raffaele and Kamoun, 2012; Rouxel et al., 2011; Spanu et al., 2010). Since 2010, whole or partial genome sequence information of additional pathogen strains, isolates and/or races with specific biological properties has emerged. By exploring sequence variation within a species, further biological insights can be gained. To date, this latter approach has been used to study pathogen effectors, but could readily be applied to determine intrinsic or acquired variation in response to novel and/or existing chemistry.

We discuss the current and potential use of large‐scale genomic resources to develop new fungicides, including the identification of prospective fungicide targets and the tools available for their validation (Fig. 1). We explore the types of studies being carried out to determine the precise mode of action of existing chemistry and the benefits for future chemical design and/or the editing of existing designs. We also consider how genome‐wide analyses can be used to discover the genes and metabolic processes involved in fungicide adaptation and resistance. Finally, we explore some of the prospective uses to which multiple fungal genome information can be applied to enhance chemical crop protection in the future.

Figure 1.

Current impact of genomics on the fungicide development pipeline.

Impact of Genomics on Fungicide Development

Most fungicides currently marketed were discovered by empirical screening—the testing of a diverse library of chemicals against target organisms, often in glasshouse tests on whole plants. This approach has the advantage of identifying chemicals with good activity, and good mobility, able to cross the cuticle, enter the plant tissue and act within the plant. Whether the traditional approach will be superseded in the near future by genome‐based technologies is unknown, although the targeting of conserved gene products essential for life in the most important pathogens is now simpler (Wang et al., 2011a).

Identification of potential fungicide targets

Various genome sequencing projects have revealed the genetic complement required for fungi to exploit different ecological niches: for example, the expansion of genes encoding proteins involved in plant cell wall degradation in the necrotrophs Botrytis cinerea and Sclerotinia sclerotiorum (Amselem et al., 2011) and the contraction of gene families encoding these enzymes in the exclusively biotrophic mildews and rusts (Duplessis et al., 2011; Spanu et al., 2010) and in some hemibiotrophs, including Mycosphaerella graminicola (Goodwin et al., 2011). Allied with genome‐wide transcriptional studies using microarrays, serial analysis of gene expression (SAGE) and, more recently, next‐generation sequencing technologies (Tan et al., 2009), genes potentially important for differentiation and pathogenicity in individual species can easily be identified and their in planta expression followed. For example, the genes required for appressorium formation in Magnaporthe oryzae (Oh et al., 2008) and for epidermal penetration and colonization in Blumeria graminis (Both et al., 2005) were identified by differential gene expression. However, the primary criterion for selection of a fungicide or antifungal compound is activity against a number of target species (so‐called broad‐spectrum activity), with no (or, at worst, minimal) effect on the host. Consequently, comparative genomics is a useful tool to identify potential targets shared by different plant pathogenic species, particularly as the number of available fungal genomes continues to increase (Table 1) and is tracked in the Genomes On Line Database (GOLD) (Liolios et al., 2010).

Comparative genomics has been used to identify novel antibacterial targets (White and Kell, 2004) and novel targets against nematodes (Kumar et al., 2007) and flatworms (Caffrey et al., 2009). In human pathogenic species, by using homologues of genes previously confirmed as essential for life in C. albicans and A. fumigatus, the sequenced genomes of eight fungi were manually mined for potential new antifungal therapies. This approach identified 10 genes present in all pathogenic fungi which are absent in humans, not auxotrophic and with accessible cellular locations. Of these, four single‐copy genes were investigated further as potential drug targets, with two (trr1, encoding thioredoxin reductase, and kre2, encoding an α‐1,2‐mannosyltransferase) sufficiently similar to Protein Data Bank templates for homology modelling. These models are now used to screen virtual chemical libraries (Abadio et al., 2011).

Comparative genomics has been used to explore the evolution of eukaryotic microbial pathogenesis, identifying sets of genes expanded in plant pathogens in comparison with free‐living fungal species, which may represent gene families with important roles in plant disease (Soanes et al., 2008), and therefore potential fungicide targets. The creation of multispecies databases, for example the Pathogen–Host Interactions Database (PHI‐base, http://www.phi‐base.org), to catalogue genes experimentally confirmed as pathogenicity genes, virulence genes, effectors or those not involved in pathogenicity, as well as fungicide targets, has greatly simplified cross‐species comparison, further enriching the comparative genomics approaches for the identification of novel targets (Winnenburg et al., 2008). For example, in an early study, 61 Fusarium graminearum genes in PHI‐base, verified as having a role in pathogenicity or virulence, were combined with gene function data from other plant‐ and animal‐infecting species. Through comparative analysis, this identified 211 homologues of pathogenicity and virulence genes in the F. graminearum genome that could be further explored experimentally as a prioritized list (Antoniw et al., 2011). The recent proposal to catalogue the properties of bioactive compounds generated by commercial and academic groups in a database, so‐called Minimum Information About a Bioactive Entity (MIABE; Orchard et al., 2011), in a format that is amenable to efficient data mining, provides a new pathway to understand how current chemistry can be improved, and, when combined with gene sequence to phenotype databases, for example PHI‐base, should facilitate the re‐optimization of the biological properties of existing fungicides, and lead to the identification of new activities. These resources will also facilitate the design of molecules active against resistant individuals (Frey et al., 2010).

Target validation

Concomitant with the increase in the available sequenced genomes of plant pathogens, there have been advances in technologies to experimentally validate genes considered as fungicide targets. These reverse genetics approaches have become higher throughput, benefiting from the immediate availability of flanking sequences and information on gene copy number, combined with advances in binary vector design and construction (for a review, see Frandsen, 2011). In addition, targeting efficiency has improved in several plant pathogenic species by the use of mutants with defects in the nonhomologous end‐joining (NEHJ) system, such as ku70 and ku80 mutants (van Attikum et al., 2001), enabling the rapid validation of gene function through gene deletion and replacement. Fungi for which NEHJ mutants are now available include the important plant pathogens B. cinerea (Choquer et al., 2008), M. oryzae (Villalba et al., 2008), S. sclerotiorum (Levy et al., 2008), My. graminicola (Bowler et al., 2010) and Alternaria alternata (Wang et al., 2011b), and the biocontrol fungus Trichoderma virens (Catalano et al., 2011). Therefore, the number of genes confirmed in different pathosystems as being crucial in infection and, in the case of T. virens, antifungal activity, now runs in the range of hundreds to thousands for any given species.

Recently, a new genomic resource available at http://www.phytopathdb.org has been developed that displays directly, within the genome browser, the exact experimentally determined in planta phenotypes for specific gene sequences when deleted or disrupted. This new resource is specifically focused on plant pathogenic species that regularly inflict serious levels of disease on globally important crop plant species, such as rice, wheat, maize, potato and oilseed rape. This gene‐centric approach could be used to link different information potentially important for fungicide target validation. However, it should be noted that the perturbation of an essential biosynthetic pathway leading to minimal growth in vitro, can sometimes be overcome during host infection if the missing metabolite(s) can be supplied from the host tissue. The latter type of failure has been noted for human‐ and plant‐infecting pathogens. For example, the enzyme ornithine decarboxylase (ODC) is essential for polyamine biosynthesis, and hence fungal growth, in vitro. However, when odc mutants of C. albicans (Herrero et al., 1999) or Tapesia yallundae (Mueller et al., 2001) infect, the host tissue supplies this missing metabolite and growth is no longer compromised, whereas, for other pathogenic fungal species, this fungicide target is readily exploitable because only low levels of host‐derived polyamines are present in the infected tissues. For example, odc mutants of the toxin‐producing necrotroph Stagonospora nodorum, which infects wheat (Bailey et al., 2000), and the biotroph Ustilago maydis, which infects maize (Valdés‐Santiago et al., 2010), cause only minimal disease.

Use of Genomic Tools to Determine the Mode of Action of the Fungicide

A consequence of the success of in vivo empirical screening of libraries for antifungal discovery is the unknown mode of action of many commercial chemicals. Although not a requirement for registration of a novel compound, the determination of the biochemical target is desirable as it enables structure–activity studies, and therefore the optimization of a compound's affinity and selectivity, as well as the use of modern polymerase chain reaction (PCR)‐based tools to monitor resistance (Fraaije et al., 2005, 2007).

Determination of the primary mode of action

Genome‐wide gene expression studies have been particularly successful in determining the physiological and metabolic consequences of antifungal chemicals. These so‐called biological spectral analyses are a useful approach for defining the similarities of the temporal responses to compounds of unknown and known activity (Fliri et al., 2005) and therefore for identifying potential new modes of action. However, the specific biochemical targets of novel compounds are more elusive. For example, the specific up‐regulation of genes involved in iron uptake and metabolism, together with in vitro phenotyping, demonstrated that exogenous iron levels are critical for ciclopirox inhibition of C. albicans (Sigle et al., 2005). Transcriptome profiling of the response of Phytophthora infestans to the novel oomycete fungicide fluopicolide, a benzamide (acylpicolide), identified genes linked to the endoplasmic reticulum and Golgi function, and, in accordance with microscopy studies, effects on vesicle transport and interactions with spectrin‐like proteins are suggested as modes of inhibition (Toquin et al., 2010). However, despite numerous transcriptomic studies, the ciclopirox (Almeida et al., 2007) and fluopicolide targets remain unknown.

Chemogenomic and chemical genetic screens are both powerful ways to determine the bioactivity of small molecules, including antifungal drugs and commercially available fungicides (Hoon et al., 2008). The approach most commonly employed in antifungal target identification and/or verification is haploid insufficiency profiling (HIP; also referred to as the ‘Fitness Test’). This exploits the diploid state of the model yeast Saccharomyces cerevisiae to determine the activity of small molecules by screening against a large collection, often genome‐wide, of heterozygous deletion strains. Strains affected are sensitized to compounds that inhibit the heterozygous locus (Giaever et al., 2004). This can identify specific small‐molecule protein interactions and, as the approach is unbiased, can capture interactions with other proteins, so‐called secondary or off‐target activity. For the human‐infecting pathogen C. albicans, a HIP or Fitness Test (CaFT), adapted from the yeast system, has been developed and used to determine the mechanism of action of novel inhibitors of microtubule formation (Xu et al., 2007) and purine metabolism (Rodriguez‐Suarez et al., 2007). The latter study identified guanine 5′‐monophosphate synthetase as the protein target. More recently, a similar approach has uncovered a new C. albicans specific antifungal target, Tfp1p, a regulator of vesicle pH (Oh et al., 2010).

In plant pathogenic fungal species, HIP studies could also be used to investigate the mode of action of fungicides. Although this approach would require a predominantly diploid species with a genome‐wide gene deletion library available, for some fungi, for example U. maydis, and oomycetes, such as P. infestans, if the appropriate investments were made, an in vitro chemical screen could be particularly informative. In vitro mutagenesis studies offer a powerful method to identify and/or further define the primary biochemical target of a compound. For unknown targets, next‐generation sequencing technologies are currently being exploited to screen resistant mutants and to define the mode of action of fungicides, although, again, this approach is dependent on a tractable sexual stage ( Box 1). In addition, the generation of mutants resistant to compounds with known targets can clarify ligand–protein interactions (Fraaije et al., 2012), and thereby provide valuable information for the rational design of more active chemicals.

1.

‘We mapped a factor involved in decreased sensitivity to a novel research antifungal using forward genetics in Mycosphaerella graminicola. From a cross between wild‐type and a less sensitive mutant, the trait was determined as monogenic. We then performed genomic high throughput sequencing – bulk segregation analysis using 30 sensitive and 30 resistant progeny which led us to a rough identification of the locus (∼340 kb). Fine mapping using CAPS markers on the whole progeny (239 isolates) identified a 12 kb mapping window where the gene responsible for the phenotype was identified. We then validated the mechanism using reverse genetics. This overall process took about a year, we are now much quicker.’ Gabriel Scalliet, Syngenta .

In haploid filamentous fungi, the systematic identification of antifungal targets is difficult as they are, by definition, essential genes. In A. fumigatus, for example, transposon mutagenesis, together with parasexual genetics, has been used for unbiased searches for essential genes (Firon et al., 2003). In addition, the predicted orthologues of known essential genes or antifungal targets, identified by HIP of C. albicans and Sa. cerevisiae, have been studied by promoter replacement strategies (Hu et al., 2007; Romero et al., 2003) or RNA interference (RNAi) (Mouyna et al., 2004). Most ascomycete fungi have the molecular machinery required for RNA silencing, with ‘knockdown’ of pathogenicity‐related genes successfully demonstrated in M. oryzae (Nakayashiki, 2005). Screens for essential for life genes in plant pathogenic fungi have not been reported, although these studies are underway. Alternatively, large‐scale reverse genetics experiments targeted to explore the function of a protein type can lead indirectly to the identification of essential for life genes. For example, in F. graminearum, the individual deletion of 116 predicted protein kinase genes identified eight causing a reduction by more than 20% in transformant growth compared with the wild‐type, whereas, for 20 genes, no transformants were recovered, suggesting that these are essential for life (Wang et al., 2011a).

Identification of multisite or nonspecific effects

In fungi, genomic studies of the fungicide response often suggest oxidative stress and the accumulation of reactive oxygen species (ROS) as a common mechanism of action. In the yeasts, Sa. cerevisiae transcripts of oxidative stress response genes are more abundant after azole treatment (Bammert and Fostel, 2000), and, in C. albicans the oxidative stress response appears to be integral to azole adaptation and, eventually, resistance (see ‘Determination of the primary mode of action’ section) (Rogers and Barker, 2003). Treatment with caspfungin, a β‐1,3‐glucan synthesis inhibitor, and fenpropimorph, an ergosterol biosynthesis inhibitor, induces the expression of genes involved in ROS scavenging and oxidative stress in the filamentous fungus Aspergillus niger, although the authors suggest that this is part of a response to morphological changes caused by the fungicides (Meyer et al., 2007). The analysis of proteome changes in C. albicans after treatment with three unrelated antifungal classes, namely polyene, azole and enchinocandin, identified a group of genes involved in the tricarboxylic acid (TCA) cycle as responsive (Hoehamer et al., 2010). A similar response after epoxiconazole treatment has been reported in the plant pathogen My. graminicola (Cools et al., 2007), with the specific up‐regulation of genes involved in mitochondrial electron transport. Epoxiconazole treatment also increases hydrogen peroxide (H₂O₂) production in healthy wheat plants, thereby enhancing the anti‐oxidative potential of plants, whilst delaying senescence (Wu and von Tiedemann, 2001). In contrast, treatment with sublethal concentrations of the triazole, prothioconazole, induced deoxynivalenol (DON) mycotoxin production in the wheat pathogen F. graminearum. The enhanced production of DON was preceded directly by a hyperinduction of H₂O₂ (Audenaert et al., 2010). These authors suggest that the inhibition of ergosterol biosynthesis increases membrane permeability, releasing H₂O₂, which, in turn, activates DON production. However, considering the evidence accumulated from global analyses of the response to azoles and other fungicides, together with the rapid induction of H₂O₂ production (4 h after treatment), in the latter F. graminearum study, a secondary nonspecific activity seems more likely.

Impact on nontarget organisms

The combination of genomics and environmental toxicology has led to the emergence of a new scientific field, termed ‘toxicogenomics’ (Nuwaysir et al., 1999). As concerns over the environmental impact of agricultural fungicide use grow, the value of genomics approaches in defining pathways of toxicity common to different nontarget beneficial organisms has also increased. Model nontarget organisms used in these studies include bacteria (Kim and Gu, 2007), yeasts (Yasokawa and Iwahashi, 2010), fish (Skolness et al., 2011) and rodents (Goetz and Dix, 2009). Recent toxicogenomics studies focusing on commercially available azole fungicides have identified common pathways affected in vitro, including genes involved in androgen and oestrogen metabolism, suggesting potential endocrine disruptor activity (Goetz and Dix, 2009). As a regulatory framework is put in place to provide guidance for the interpretation of these genome‐wide data (Goetz et al., 2011), these studies are likely to become an increasingly important component of European pesticide legislation.

Use of Genomics to Explore Fungicide Resistance

Before the era of genomics, studies of the genetic mechanisms of adaptation and resistance to fungicides in fungal pathogens were based on classical genetics allied with biochemical and candidate gene approaches. High‐efficiency transformation protocols to confirm the impact of genes suspected to confer resistance have been available for a few plant pathogens and saprophytes for some time (Steffens et al., 1996). Through the use of plasmids for homologous gene replacement and/or heterologous expression (Gems et al., 1991; Tsukuda et al., 1988), early descriptions of the molecular mechanisms underlying resistance were forthcoming. For example, resistance mechanisms were reported for the methylbenzimidazole carbamates (MBCs) (Jung et al., 1992) and the succinate dehydrogenase inhibitor carboxin (Keon et al., 1991). These early studies were dominated by the tractable species Aspergillus nidulans and U. maydis.

Recent studies of resistance to novel fungicides use the same principles, although the newer next‐generation sequencing technologies can substantially enhance the process. For most fungicide classes, field resistance is conferred by point mutations in the target gene encoding an amino acid change that prevents fungicide inhibition without perturbing the activity of the target protein (Brent and Hollomon, 2007). Consequently, sequencing of target genes has proved to be successful in identifying the variant sequences responsible for resistance when a clear relationship to phenotype can be established: for example, by revealing the mutations in the β‐tubulin gene causing resistance to MBCs (for a review, see Ma and Michailides, 2005) and in the cytochrome b gene causing resistance to the quinone outside inhibitors (QoIs; for a review, see Fernandez‐Ortuno et al., 2008).

When target site changes are absent or their effect on phenotype is uncertain, or classical genetics has identified more than one gene associated with a resistance phenotype, the advantages of using an unbiased genome‐wide experimental approach to identify altered genes, transcripts, proteins and/or metabolites in a fungicide‐resistant individual are clear, for example in the analysis of azole resistance mechanisms in plant and human pathogenic fungi (Becher et al., 2011; Cools et al., 2007; De Backer et al., 2001; Ferreira et al., 2006; Silva et al., 2011). Perhaps not so apparent, however, is the power of using genomics to reveal both the conserved and divergent molecular response pathways involved in the adaptation and development of resistance to different fungicide classes. These types of studies will, over time, reveal whether a conserved response to synthetically generated xenobiotics exists, and whether this is the same or different to common responses induced by a range of natural xenobiotics.

Use of genomics to identify the mechanisms of resistance to fungicides

For several human pathogenic fungi, the analysis of transcriptional changes underlying their adaptation to unrelated fungicides, eventually leading to stable resistance, has permitted the characterization of common programmes of evolution (Cowen et al., 2002). These changes are dependent on a set of co‐regulated genes (Sanglard et al., 2009). For example, the ATP binding cassette (ABC) transporters CDR1 and CDR2 (Candida drug resistance 1 and 2) and the major facilitator gene MDR1 (multidrug resistance 1) are commonly overexpressed in clinical and experimentally evolved azole‐resistant strains of C. albicans (Cowen et al., 2002; Sanglard et al., 1995; White, 1997). Several genes coordinately regulated with both CDR genes and MDR1 are also responsive to the β‐tubulin inhibitor benomyl, and are up‐regulated by H₂O₂ (Enjalbert et al., 2003). As mentioned previously, genome‐wide transcriptional profiling has suggested that oxidative stress is a general mechanism of fungicide action. However, the discovery of a transcriptional regulator of these genes, MRR1 (multidrug resistance regulator), provides a mechanism for concurrent responses to oxidative stress and fungicides (Cowen et al., 2002) which can mediate the development of fungicide resistance (Rogers and Barker, 2003). Indeed, clinical isolates of C. albicans resistant to multiple unrelated antifungals have been shown to carry gain‐of‐function mutations in MRR1 that result in constitutive overexpression of MDR1, as well as genes involved in the regulation of the cellular redox state (Karababa et al., 2004; Morschhauser et al., 2007). Transcriptional profiling of Candida parapsilosis strains correlated resistance to fluconazole and voriconazole with MDR1 overexpression. The resistance phenotype was again caused by point mutations in MRR1 (Silva et al., 2011).

Similarly, recent studies of strains of the plant pathogen B. cinerea, resistant to a variety of unrelated fungicides, identified activating mutations in the orthologous transcription factor (MRR1), controlling the expression of the ABC transporter AtrB (Kretschmer et al., 2009). ABC transporters have also been implicated in resistance to unrelated fungicides in M. oryzae (Lee et al., 2005), Penicillium digitatum (Nakaune et al., 1998), Oculimacula yallundae and My. graminicola (Leroux and Walker, 2011). Although a multidrug‐resistant phenotype in My. graminicola has not been reported in other studies, and microarray analysis failed to identify an ABC transporter responsive to fungicide treatment (Cools et al., 2007).

Use of genomics to explore intrinsic resistance to fungicides

Several important human and plant pathogenic fungi are unaffected by some of the most widely used broad‐spectrum fungicides. For example, A. fumigatus is intrinsically resistant to the azole fungicides fluconazole and ketoconazole, although inhibited by voriconazole and itraconazole. Strains of the globally important soil‐dwelling pathogen, Gaeumannomyces graminis, which causes take‐all disease in many small‐grain cereal species, including wheat and barley, can exhibit intrinsic resistance to silthiofam, which leads to the inhibition of mitochondrial ATP transport (Joseph‐Horne et al., 2000). Several species of Fusarium are naturally resistant to some azoles. Fusarium solani (syn. Nectria haematococca), for example, an opportunistic pathogen of both humans and plants, is resistant to all commonly used medical azoles, including fluconazole, itraconazole, voriconazole and the recently introduced posaconazole (Nucci and Anaissie, 2007). The analysis of genome sequences now publicly available for four Fusarium species has identified multiple copies of the CYP51 gene encoding the azole target sterol 14α‐demethylase (Becher et al., 2011), contradicting the long‐held view that CYP51 exists as a single gene in all phyla. Indeed, genome information has revealed that paralogous CYP51 genes are common in ascomycete fungi, including species of Aspergillus, the rice blast fungus M. oryzae and the wheat pathogen Pyrenophora tritici‐repentis. In microarray experiments, the CYP51A gene of A. fumigatus and F. graminearum is consistently the most transcriptionally responsive gene to disruption of ergosterol function (Becher et al., 2011; Gautam et al., 2008; Liu et al., 2010). In addition, CYP51A disruptants in A. fumigatus (Mellado et al., 2005), M. oryzae (Yan et al., 2011) and F. graminearum (Liu et al., 2011a) are more sensitive to azoles, confirming the importance of CYP51A in intrinsic resistance. Unlike other plant pathogenic Ascomycete species, F. graminearum is also intrinsically resistant to amine fungicides. These are ergosterol biosynthesis inhibitors that inhibit at C‐14 reductase, encoded by ERG24. Genomic analysis has identified two ERG24 genes (ERG24A and ERG24B; Liu et al., 2011b) on different chromosomes. Therefore, recent genome information has provided a rationale for the ineffectiveness of some broad‐spectrum fungicide classes against important pathogens. This is vital information for the development of new agrochemicals. Conversely, gene sequence information can indicate potential limitations to resistance development in particular pathogens. For example, a number of species of rust and Alternaria have a self‐splicing type 1 intron in the mitochondrial cytochrome b sequence located directly after codon 143. This prevents the evolution of QoI resistance conferring G134A substitution, as interruption of intron splicing would inactivate the protein. Therefore, widespread QoI resistance is unlikely to occur in rusts or Alternaria spp. (Grasso et al., 2006).

Use of genomics to identify chromosomal changes conferring fungicide resistance

Comparative genomic hybridization, a microarray‐based technique to analyse gene copy number across the genome, has identified karyotype variability as a common mechanism of antifungal resistance in C. albicans, with aneuploidy, primarily trisomy of chromosome 5, more common in fluconazole‐resistant strains (Selmecki et al., 2006) than in sensitive strains. The phenotype is conferred by the amplification of ERG11 (CYP51) and transcriptional activator of CDR genes 1 (TAC1), a regulator of ABC transporter expression, located on the long arm of chromosome 5 (Selmecki et al., 2008). Similar chromosomal rearrangements are correlated with azole resistance in Cryptococcus neoformans, with duplication of chromosome 1, on which the CYP51 and an ABC transporter gene antifungal resistance 1 (AFR1) reside, associated with fluconazole adaptation (Sionov et al., 2010). The contribution of karyotype variation to resistance to synthetic fungicides in plant pathogens has not been reported, although, in F. solani isolates, resistance to the plant‐derived antimicrobial phytoalexin pisatin was mapped to a mitotically unstable chromosome dispensable for normal growth (Coleman et al., 2009; Miao et al., 1991). Indeed, a number of plant pathogens have dispensable chromosomes (Covert, 1998), some of which are known to cause phenotypic variation (Hatta et al., 2002), whereas others, for example in My. graminicola, have no clear function (Goodwin et al., 2011). In the future, genome‐wide analyses will undoubtedly elucidate a role for chromosome variation in fungicide resistance.

Outlook

Within less than 10 years, the fungal research community has moved from a position in which a reference genome was available for just a handful of human and plant pathogenic fungal species to a position today in which a reference and annotated genome sequence is available for hundreds of pathogenic and nonpathogenic species across the fungal tree of life. This new knowledge is set to have an impact on every part of agrochemical development (Fig. 1).

Impact on discovery

Where the use of genomic information is currently not part of the discovery toolkit, it soon will be. For example, the percentage of proteins annotated from the genome as hypothetical proteins or conserved hypothetical proteins in a given species is typically in the range 40%–55%, although in some of the newly sequenced obligate biotrophs, e.g. rusts and mildews, this figure can rise to over 80% (Cantu et al., 2011; Duplessis et al., 2011; Spanu et al., 2010). This large subset of predicted proteins represents the untapped potential for chemical intervention. By exploring their expression and abundance patterns in a range of different pathogenic species in response to the same chemical class, and thereby linking the subsets of conserved hypothetical sequences across species, novel chemical targets can be derived. Currently, most fungicides target either cytoplasmically localized or membrane‐bound proteins. For the genes annotated as coding for conserved hypothetical or hypothetical proteins, various bioinformatic software can be used to predict their potential cellular location [for example, ProtCom (ProtComp v8.0; http://www.softberry.com) or WolfPsort (Horton et al., 2007)], and this will provide an initial layer of annotation. Second, some predicted hypothetical proteins are now recognized to contain a specific domain of unknown function (DUF) or computationally derived motifs. These new types of sequence‐derived information can be used to further annotate existing pathogen chemical induction transcriptomics and proteomics datasets based on the putative cellular location of the predicted hypothetical protein. A further refinement of the list of hypothetical proteins that could be explored for fungicide targeting is to determine the gene copy number across multiple species. For example, a single‐copy highly conserved sequence across species could be ideal as a potential multispecies target, whereas the discovery of variable copy numbers and high copy numbers across species may immediately raise concerns.

An emerging new technical advance is host‐induced gene silencing (HIGS). In HIGS, gene silencing is triggered directly from the host plant and is used to target genes in the plant pathogen or pest. Already, HIGS has been successfully applied to study gene function in two cereal‐infecting obligate biotrophs: Blumeria graminis f. sp. hordei during the infection of barley leaves (Nowara et al., 2010) and Puccinia striiformis f. sp. tritici during the infection of wheat leaves (Yin et al., 2011). The triggers for these trans‐species gene silencing events are short double‐stranded RNAs (dsRNAs); however, the mechanism(s) involved in delivering the RNA silencing signals from the plant cell into fungal cells is not yet known (reviewed in Lee et al. (2012). For obligate biotrophic fungal species, the availability of this exciting new tool should permit the development of a novel in planta route for the discovery of suitable intervention targets. In theory, HIGS could also be used directly to deliver RNA silencing‐based products against the target gene, and thereby to directly control pathogen and pest infection. This approach would require HIGS to be highly efficient within the growing crop. A control strategy based on RNA silencing in the pathogen would obviate the need to identify chemicals that interfere with either a newly discovered target or an existing intervention target.

Impact on lead optimization

Screening of activities within libraries of newly discovered agrochemicals using ‘omic’ (transcriptomic, proteomic, metabolomics) technologies, together with the improvement in in silico modelling tools, will be facilitated by the ever‐increasing numbers of different plant pathogen genomes available and, moreover, the genomes of specific individuals within a species which exhibit distinct biological properties. Collectively, these data will greatly enhance the detection and optimization of broad‐spectrum inter‐ and intraspecific activity. The integration of genome information from different pathogens and individuals within populations can be used to select additional common target sites present in a broad range of problem species. In addition, with the definition of potential detoxification pathways, for example expanded P450‐omes of ascomycete fungi (Cresnar and Petric, 2011), our understanding of fungicide degradation will be enhanced, enabling the modification of new chemicals that would previously have been discarded. These attractive new targets for chemical intervention could improve overall existing product efficacy and longevity, or could be used to start the discovery pipeline afresh.

Impact on registration

Once a regulatory framework is in place, toxicogenomic profiling of a newly discovered and optimized fungicide chemistry will rapidly provide information on the environmental consequences of the use of a fungicide, thereby circumventing the need for animal models, and speeding up the registration process. An assessment of resistance risk would also be enriched by profiling the genomic response of a target pathogen.

Impact on mode of action studies

To date, the impact of genomics on novel mode of action studies has been minimal, primarily because of the difficulty in identifying the primary biochemical target. Most detected biological effects are the secondary consequences of fungicide application leading to restricted pathogen growth. However, in the future, by combining labelled chemistry with transcriptomics and/or proteomics, early cellular distribution and response can be explored simultaneously, thereby potentially providing the subcellular locations of targets. In addition, the screening of mutagenized pathogen populations for strains with small shifts in sensitivity, followed by genome resequencing of multiple strains, may reveal the target sequences of potential interest.

Impact on stewardship

For the commercial launch of a novel chemistry, genomic information will increasingly be used to assist in prolonging the longer term efficacy of each chemical class. Advances in sequencing technologies and the reduction in costs mean that, in the very near future, genome‐wide information for a species or isolate found at a new location will be available within days or even hours of initial discovery. Shortly afterwards, the sequence variation will have been computationally predicted and available for scrutiny via a genome browser. By monitoring simultaneously the global changes to a genome, i.e. the shifts in the entire ‘genome landscape’, as well as the local target site‐specific changes occurring in problematic pathogenic species within intensive production systems, it should be possible to re‐use existing target sites by modelling the changed chemical sites, and thereby to design novel chemistry that can intervene.

Conclusions

The availability of genomic sequence information for an increasing number of pathogenic species and isolates with distinct lifestyles and unique biological properties now provides all aspects of fungicide research with a comprehensive and informative starting point. With the financial return on the efficient production of healthy food predicted to rise, there is a keen interest in the development of innovative ways to combine the plethora of genomic information with advances in modelling and chemical synthesis, and to align this with our ever‐increasing understanding of pathogen biology and the pathogenicity process. Undoubtedly, the future rewards to fungicide research achievable through the application of genomic and post‐genomic technologies will be both numerous and exciting. However, although full genome sequencing and transcriptomic studies will be able to provide, within a few days, deep insight into the evolutionary changes that have recently occurred in important pathogen populations and, via modelling, potential ways to modify chemistry to regain the upper hand, some obstacles remain. New chemistry tailored to novel targets will still have to pass the acid tests of activity in planta and efficacy in the field. Furthermore, as it stands, both new and modified chemicals are required to complete the full registration process that normally takes several years. Potential markets and new chemical solutions for problem diseases will not be accessed immediately. Therefore, unless a faster route to register chemistry from a preranked list of the potential safest chemical compounds can occur, possibly based on prior knowledge, the full effects of genomic devised solutions to intervention and pathogen control in a commercial crop setting and/or clinical practice will, at best, remain slow. Nonetheless, these new approaches and technologies promise to accelerate the rate at which novel candidates are identified and to improve the breadth of the product pipeline.