You are on Érudit's new platform. Enjoy! Switch to classic view

Applying comparative genomics to plant disease epidemiology

  • Linda M. Kohn

…more information

  • Linda M. Kohn
    Department of Botany
    University of Toronto at Mississauga
    Ontario L5L 1C6


Article body

Taken broadly, the term “comparative genomics” usually conveys parallel study of genome structure, dynamics, and function among different individuals or groups of individuals (taxa) at the different ranks in the taxonomic hierarchy, from species to kingdom. Genome comparisons at different phylogenetic distances answer different sorts of questions (Hardison 2003). The reader is directed to Figure 1 in Hardison at =index-html. Short distances are defined by a small number of steps, such as nucleotide substitutions or changes in microsatellite repeats, often associated with recent evolution on a contemporary timescale. In a branching phylogenetic tree, these patterns are usually represented as the tip branches (branches are also termed clades). In contrast, long distances are defined by a large number of steps, associated either with more ancient ancestry or with accelerated evolutionary rates. In a phylogeny, these are often represented on the internal branches.

DNA sequence-based studies of single or multiple genomic regions can delimit species or the patterns of gene flow that define populations of a species. While such studies seek genomic differences, comparative genomics often targets similarities among genomes to identify the genes that are relatively invariant, i.e. conserved, because of their significant function. Mutational variants of genes that do not improve the fitness of the individual tend to be lost through natural selection. Mutations that change the function or protein products of genes that have important functions tend to be weeded out through a kind of natural selection termed “purifying” selection, with the result that the DNA sequences of these genes tend to be conserved rather than highly variable. On a contemporary evolutionary scale, mating type genes are conserved within species and in some fungi, such as the Fusarium graminearum complex, among closely related species (O’Donnell et al. 2004). On a more ancient scale, comparative genomics seeks the genes for core proteins common to all organisms, such as MAP kinases in protein-protein signaling or ABC transporters as efflux pumps.

All of these aspects of comparative genomics can contribute to plant disease epidemiology. Accurately defined species are key for quarantine strategies, as well as for patenting of biocontrol agents and biotechnological products or processes. Knowing the closest relatives of a species predicts biological characteristics, most notably of fungi known only in their asexual state that lack the necessary morphological features for taxonomic diagnosis among fungi with known sexual states. A good example is Sclerotium cepivorum, cause of white rot of onion and garlic, which has no known sexual state. Species of the form genus Sclerotium produce only mycelium and sclerotia. Some are asexual Basidiomycetes, e.g. S. rolfsii, and some are Ascomycetes, e.g. S. cepivorum. Based on phylogenies of seven genes or genomic regions, there is substantial evidence that S. cepivorum is closely related to the genus Sclerotinia (Carbone and Kohn 1993; Carbone 2000; Holst-Jensen et al. 2004). Control strategies proven to be effective for sexual relatives have potential for such asexual orphan species. As more fungal plant pathogens are fully sequenced methods in comparative genomics extended to closely related species will fully exploit the resource information on gene regulation of pathogen development, pathogenesis, and mechanisms of virulence (Yarden et al. 2003).

Recently, we have traced the distribution of avirulence (AVR) genes in genotypes of the rice blast agent, Magnaporthe oryzae (of the M. grisea species complex), against an evolutionary framework that provides information on geographical and host distribution (Couch 2004). With powerful statistical approaches, such as coalescence, Bayesian, and statistical parsimony, comparative genomic data can be explored to address basic questions on the origins, movement and amplification of epidemiologically significant genotypes (Carbone and Kohn 2001, 2004; Milgroom and Peever 2003; Phillips et al. 2002).

Phylogenetic or genealogical interpretation of DNA sequence data from multiple genomic regions has become the gold standard for species delimitation and population genetics. The basic unit of analysis is the haplotype. All members of a haplotype share the same variable nucleotide sites in a sequence, as well as all of the invariant sites. A species defined by the multilocus standard is a clade based on concordant DNA sequences from multiple genes or genomic regions, termed loci, corroborated if possible by proof of interfertility among individuals within the clade and lack of fertility with individuals from other clades. The genomic part of this approach has been called the genealogical concordance phylogenetic species recognition (GCPSR) concept (Taylor etal. 2000; O’Donnell et al. 2004). In the absence of the interfertility criterion, the GCPSR concept may not distinguish population divergence from speciation, e.g. individuals in the same species but in two evolutionarily diverging populations would be expected to be able to mate but could easily be represented in a phylogeny as two distinct tip clades (Carbone and Kohn 2001). Species and populations within species can be distinguished by sampling of more than one population within each putative species, then screening sequences of multiple loci. The screen should include some loci that are highly polymorphic, variable within population samples of a single species, i.e. at the contemporary time scale, as well as loci that are invariant within populations, but variable among species, i.e. at the more ancient time scale. With this type of sample and dataset, species and divergent populations can be distinguished, even in the absence of interfertility tests. This approach also has the advantage of removing sampling bias when only a few isolates are studied for each species pre-determined to be of interest (Hare 2001). For rapid throughput of large samples, we initially sequence a small set of isolates (usually 12) that are likely to vary. Once polymorphic sites are identified, samples are screened for nucleotide variation by means of single-strand conformation polymorphisms (SSCPs; Carbone et al. 1999).

Comprehensive multi-locus sequencing of large samples reveals several important aspects of molecular evolution in fungi. In recent studies in our laboratory, we screened 383 isolates of Sclerotinia sclerotiorum and related species (Carbone and Kohn 2001) and 497 isolates of Magnaporthe oryzae (Couch 2004) plus isolates of additional species (Couch and Kohn 2002). First, in both studies, it is evident that one gene or genomic region may be invariant at the population level but highly polymorphic among isolates representing different species, e.g., the MPG1 gene in the M. grisea complex with 12 polymorphic sites among population samples of M. oryzae, which is associated with rice and other grasses excluding Digittaria (crabgrass), and 49 polymorphic sites in the species complex that includes M. oryzae, and what is now recognized as M. grisea, which is associated only with crabgrass ( Couch 2004; Couch and Kohn 2002). This feature is also evident in Sclerotinia, in which some loci yield many haplotypes and branches within species but short branches separating species, while other loci yield only one or two haplotypes within species but long branches separating species. For example, the intergenic spacer of the nuclear rDNA repeat (IGS), with 55 informative sites yields 47 haplotypes within S. sclerotiorum, offers solid evidence of population divergence within the species. In species other than S. sclerotiorum, the IGS was present in more than one form as highly divergent paralogues. Consequently for this level of comparison the promoter region of the IGS was used, yielding 18 haplotypes and distinguishing all species of Sclerotinia on relatively long branches. Louisiana and Ontario populations of a Trillium parasite could not be distinguished but this taxon and Dumontinia and S. cepivorum formed a clade. In contrast, the Actin gene with 14 haplotypes based on 44 informative sites, yielded only one haplotype, based on one informative site, in S. sclerotiorum. However at species level it clearly distinguishes on long branches all four species of Sclerotinia (S. minor, S. sclerotiorum, S. trifoliorum, and one new species from Norwegian potatoes, wild Caltha palustris (marsh marigold) and Taraxacum (dandelion). In the Actin trees, the Dumontinia clade, including S. cepivorum, was marked by shorter branches (i.e. fewer differences as evidenced by nucleotide substitutions) than in the IGS tree.

A second aspect of molecular evolution is that evolutionary rates may vary among loci, as shown by different branch lengths in phylogenies for one sample of isolates in the Sclerotiniaceae. This point is well illustrated in comparing branch lengths among species within Sclerotinia and among species representing other genera of the Sclerotiniaceae for the IGS and Actin loci, as well as the calmodulin, RAS protein, chitin synthase, and the translation elongation factor-1 alpha (TEF or EF1-alpha) (Carbone 2000; Carbone and Kohn 2001; Holst-Jensen et al. 2004). Basically, each locus may tell a different evolutionary story, as in the Sclerotiniaceae, which is why a multi-locus approach is important to approximate the true evolutionary tree. In contrast, in the M. grisea complex, there is remarkable concordance among three loci, both with respect to rates and topology (Couch and Kohn 2002).

Thirdly, the rate of molecular evolution may exceed that of morphological evolution. M. oryzae and M. grisea, two distinct biological species, are clearly demarcated phylogenetic species based on data from several loci, but present no detectable morphological difference (Couch and Kohn 2002). Despite the rather long branches separating species, most of the species of the Sclerotiniaceae produce sexual states with remarkably similar morphology.

A fourth characteristic of molecular evolution, albeit with certain caveats, is that based on multi-locus data in a contemporary sample, molecular evolution reflects pathogen evolution. Within our phylogeny of M. oryzae haplotypes, we detected evidence for one host jump from millet (Setaria) to rice with a corresponding increase in copies of a transposable element in rice haplotypes (Couch 2004), consistent with expectations for asexual populations in the absence of meiosis (when phenomena, such as repeat induced point mutation or “ripping”, can eliminate copies). As no evidence of recombination was detected among rice haplotypes, the indication is that rice populations are indeed asexual. Interestingly, jumps from rice to weeds of rice were observed. These weed haplotypes had few copies of the transposable element. They were weakly or non-pathogenic on rice, consistent with high host specificity associated with clades, as evidenced in co-inoculation experiments in collaboration with B. Valent (Kansas State University) and Didier Tharreau (Montpellier, France).

Last in this list of intriguing molecular evolutionary phenomena is phylogenetic conflict, indicative of mutational hotspots or of recombination. Rather than viewing conflict with dread as an obstacle to inferring a well-supported phylogeny, conflict within or between loci should be investigated with enthusiasm, especially at the level of divergent populations within a species. At higher taxonomic rank, problems with inferring a phylogeny can also arise as a result of hybridization or gene duplication events and can be difficult to sort out. In the context of epidemiology and disease management, recombination in pathogen populations can be a quick route to building a super pathogen by combining phenotypic traits, such as virulence, fungicide resistance, or enhanced tolerance or response to environment in one genotype. On the other hand, in the absence of conflict due to recombination, clonality through asexual reproduction or self-fertilization (in a haploid organism) is often associated with epidemic spread. Conflict can be detected by a lack of resolution in phylogenies (a comb- or star-like structure for the whole phylogeny, or a fan in one or more clades), and localized to specific loci or parts of a phylogeny by a variety of methods, including compatibility matrices, partitioned Bremer support, or inference of networks rather than branching phylogenies (Carbone et al. 1999; Carbone and Kohn 2001; Posada and Crandall 2001; Posada et al. 2002). Recombination “hot-spots”, blocks of sequence that recombine with each other (with comparatively less or no recombination within a block) are an object of interest in current studies of human evolution (Pennisi 2004); recombination blocks which may or may not currently be hotspots are also found in fungi, as we have observed in M. oryzae and S. sclerotiorum (Carbone and Kohn 2001; Couch 2004). Such recombination blocks are evident in a compatibility matrix as large blocks of incompatibility within a locus and are most readily observed in large, contiguous sequences, such as the 4 Kb span of the intergenic spacer region of the nuclear rDNA repeat in S. sclerotiorum.

So what is next? As of July 1, 2004, Sclerotinia sclerotiorum will be sequenced at the Broad Institute, Massachusetts Institute of Technology, financed by the USDA as part of the NSF-USDA Microbial Genome Sequencing Program (M. Dickman, L.M. Kohn, J. Rollins). Like the Magnaporthe grisea Genome project ( ), an invaluable resource for our marker development, the Sclerotiniasclerotiorum genome will be publicly available.

Parties annexes