Pathogenicity Determinants in Smut Fungi Revealed by Genome Comparison

See allHide authors and affiliations

Science  10 Dec 2010:
Vol. 330, Issue 6010, pp. 1546-1548
DOI: 10.1126/science.1195330


Biotrophic pathogens, such as the related maize pathogenic fungi Ustilago maydis and Sporisorium reilianum, establish an intimate relationship with their hosts by secreting protein effectors. Because secreted effectors interacting with plant proteins should rapidly evolve, we identified variable genomic regions by sequencing the genome of S. reilianum and comparing it with the U. maydis genome. We detected 43 regions of low sequence conservation in otherwise well-conserved syntenic genomes. These regions primarily encode secreted effectors and include previously identified virulence clusters. By deletion analysis in U. maydis, we demonstrate a role in virulence for four previously unknown diversity regions. This highlights the power of comparative genomics of closely related species for identification of virulence determinants.

Smut fungi are biotrophic pathogens causing disease in a number of agriculturally important crop plants. Ustilago maydis and the related fungus Sporisorium reilianum both parasitize maize (1, 2). Their life cycle leading to the infectious form is similar (2, 3); however, shortly after infection U. maydis locally induces tumors on all aerial parts of the plant, whereas S. reilianum spreads systemically and causes symptoms in male and female inflorescences only (Fig. 1). Both S. reilianum and U. maydis establish an intimate communication with their host through secreted protein effectors that enable biotrophic development (3, 4). Effector proteins like U. maydis Pep1 can suppress plant defense responses (5). Additional effector genes were identified in the genome as genes encoding U. maydis–specific secreted proteins, most of which are up-regulated during host colonization (3). Many of these effector genes are clustered, and deletion of five of these clusters affected virulence in seedlings (3). Some cluster genes are induced in specific plant organs, and respective cluster mutants show altered virulence depending on the host tissue infected (6). In plant parasitic oomycetes, genes for effector proteins are under diversifying selection and occur in highly flexible genomic regions (7). In accordance with this emerging picture of plant-pathogen communication via rapidly evolving effector proteins, we hypothesized that virulence-associated U. maydis genes might be identified as genomic regions with high sequence variability in closely related smut species.

Fig. 1

Infection symptoms of U. maydis and S. reilianum. Tumor formation on ear (A), tassel (B), and leaf (C) after inflorescence infection [(A) and (B)] or seedling infection (C) by U. maydis. Maize seedling infection with S. reilianum leads to formation of spores (arrowheads) and leaflike structures (arrows) in ear (D) and tassel (E), whereas inoculated leaves show mild chlorosis but no tumors (F). Scale bars indicate 1 cm.

To identify regions of high diversity in the U. maydis genome, we sequenced the genome of S. reilianum strain SRZ2 (8). The S. reilianum genome assembly covers 97% of the 18.7-Mb genome (9). As in U. maydis (3), the genome is organized in 23 chromosomes, to which 6648 gene models could be assigned after manual annotation. The genomes of U. maydis and S. reilianum exhibit a remarkable degree of synteny (Fig. 2A) (10) despite an average amino acid identity of predicted proteins of only 74.2% (Fig. 2B). Interestingly, some chromosome ends are extended by up to 20 genes in U. maydis compared with ends in S. reilianum. About 90% of these chromosome end–associated genes do not carry any functional annotation and no enrichment for secreted effectors is evident (fig. S1), whereas the others likely encode enzymes for secondary metabolism (table S1). Because orthologs of these genes are lacking in S. reilianum, their presence is likely dispensable for virulence. Compared with an average amino acid identity of 76% for nonsecreted proteins, secreted proteins in both organisms display an average identity of only 62% and are enriched among the weakly conserved proteins (Fig. 2B). This suggests that genes coding for secreted proteins are subject to more rapid evolution.

Fig. 2

Comparison of U. maydis and S. reilianum genomes. (A) Synteny (diagonal lines) of protein-encoding genes on the 23 chromosomes of U. maydis compared to those on the 23 chromosomes of S. reilianum. (B) Distribution of amino acid identities of all protein-encoding genes occurring in both genomes (right axis, blue bars). The percentage of proteins with a predicted secretion signal is given for each amino acid identity value (left axis, red bars).

Manual sequence comparisons of predicted gene models of S. reilianum and U. maydis led to a reannotation of more than 300 gene models of U. maydis (table S2). The S. reilianum genome has a 5.7% higher GC content than the U. maydis genome and a 5% higher coding potential (table S2). More than 99% of all InterPro ( domains are equally or close to equally represented in the two genomes, suggesting that the biosynthetic repertoire of both species is comparable. However, S. reilianum contains three putative RNA-dependent RNA polymerase genes (table S3). A search for other components (11, 12) of a putative RNA interference (RNAi) machinery in S. reilianum identified homologs of dicer and argonaute (table S3). These genes all lie in highly syntenic regions (10); however, the corresponding intergenic regions in U. maydis lack traces of the respective genes. Instead, we detected between one and four variants of a conserved 10–base pair (bp) sequence highly overrepresented in intergenic regions of U. maydis (3) that is largely absent from the genome of S. reilianum (table S4). We speculate that the genes encoding components of the RNAi machinery were lost in U. maydis by rare homologous recombination events involving the 10-bp motifs. To investigate whether the generation of small regulatory RNAs could explain the differences in symptoms of U. maydis and S. reilianum, we deleted the putative dicer gene sr16838 in a solopathogenic strain of S. reilianum (8). Infection experiments revealed that sr16838 deletion mutants were affected in neither virulence nor symptom development (fig. S2).

To detect regions of high sequence divergence, we compared the genomes of U. maydis and S. reilianum gene by gene (8, 10) and identified regions encoding genes with low sequence conservation (“divergence clusters”) in a conserved genomic context (8). This analysis revealed the presence of 43 divergence clusters (table S5) (10). Seventy-one percent of the genes in divergence clusters occurred in both organisms, whereas 19% were S. reilianum–specific and 10% were U. maydis–specific. Sixty-one percent of the genes in divergence clusters are predicted to encode secreted proteins. In contrast, of all S. reilianum and U. maydis genes less than 12% encode potentially secreted proteins. Ninety-four of the genes in divergence clusters code for proteins without functional annotation (table S5). It is notable that 442 of the 494 putative effectors detected in U. maydis are conserved in S. reilianum, with amino acid identities ranging from 20 to 98% (10). In addition, 445 genes are present only in U. maydis, whereas 372 exist only in S. reilianum, and of these about 15% encode secreted proteins.

Among the divergence gene clusters identified in this study were 7 of the 12 previously described U. maydis effector gene clusters (3), and these included all clusters whose deletion affected virulence in seedling infections (3) (fig. S3). In contrast, four of the gene clusters of secreted proteins without a deletion phenotype in U. maydis seedling infections (3) did not classify as divergence clusters (fig. S4). To test whether any of the newly identified divergence gene clusters also harbor virulence factors, we individually deleted six randomly chosen divergence clusters in U. maydis (Fig. 3). In seedling infection assays, loss of three divergence gene clusters (15-12, 5-21, and 20-15) attenuated virulence, whereas deletion of two clusters (5-18 and 11-16) did not affect virulence (Fig. 3 and fig. S5). The absence of a virulence phenotype likely reflects redundancy because in these cases potential paralogs exist elsewhere in the genome (table S6). In cluster 8-12, we detected the mig1 gene that is highly induced during plant colonization and encodes an effector with similarity to apoplastic fungal avirulence proteins (13, 14). Cluster 8-12 contains three additional genes, of which two encode proteins with similarity to Mig1. In S. reilianum, the mig1 gene family is expanded to eight members residing in a single cluster (Fig. 3 and fig. S6). Whereas the deletion of only mig1 did not affect virulence (13), cluster 8-12 deletion caused hypervirulence (fig. S5). Hypervirulence could result from an active attenuation of fungal proliferation by respective effectors (3). However, given the conserved features between avirulence proteins and Mig1 effectors, we now propose that genes whose deletion leads to hypervirulence encode weak avirulence proteins that trigger defense responses in plants expressing a cognate resistance protein, resulting in an attenuation of fungal growth.

Fig. 3

Gene-by-gene comparisons of divergence clusters between U. maydis (Um) and S. reilianum (Sr) deleted in this study. (A) Cluster 15-12, (B) cluster 20-15, (C) cluster 11-16, (D) cluster 8-12, (E) cluster 5-21, and (F) cluster 5-18. Genes encoding putatively secreted proteins are shaded in gray. Bars connecting syntenic homologs are color-coded (green, high; yellow, weak conservation), and numbers give amino acid identities. Brackets denote regions deleted in U. maydis mutants. Virulence phenotypes of the respective mutants are indicated; the corresponding scores are found in fig. S5.

Although most of the deleted genes in the four divergence clusters with an effect on virulence encode putatively secreted proteins, cluster 15-12 encodes only proteins without identifiable secretion signals, suggesting that additional mechanisms for virulence modulation exist in U. maydis. With respect to the origin of these clusters, we do not detect hallmarks for horizontal gene transfer like an altered GC content or an association with repetitive elements (fig. S7). Therefore, because many divergence regions contain members of small gene families (table S4) we propose that the majority of divergence clusters have been generated by local gene duplications followed by strong natural selection resulting from interaction with different host molecules.

Our studies demonstrate the power of comparative genomics of closely related species for the identification of new virulence genes. The U. maydis and S. reilianum pathosystems present a unique example of differentiation of two closely related pathogens parasitizing the same host. We have recognized that the U. maydis and S. reilianum genomes comprise conserved effector genes as expected for pathogens infecting the same host. However, although the two pathogens are both recognized and challenged by the maize immune system, they also possess strongly differentiated effectors, suggesting that they are targeting different host molecules. We speculate that their different infection strategies lead to the interaction between different host-pathogen molecules and thereby the evolution of differentiated sets of effector proteins, although we cannot exclude contributions of the species-specific genes. The assertion that closely related pathogens interact with and affect different host targets suggests a high variety in pathogen targets within the same host. It also suggests a temporal and spatial difference in the composition of different host proteins, which can drive the evolution of different sets of effectors in pathogens with different infection strategies.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Tables S1 to S6


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We are grateful to the Max Planck Society for funding genome sequencing and annotation. T.B. was supported through the Deutsche Forschungsgemeinschaft via SFB593. A.M.M. received a Humboldt fellowship. H.G. was funded by the International Max Planck Research School for Environmental, Cellular, and Molecular Microbiology. R.K. acknowledges support through the LOEWE program of the State of Hesse. We thank U. Bonas and G. Gottschalk for critical comments on the manuscript. The sequence was submitted to European Molecular Biology Laboratory database under accession numbers FQ311430 to FQ311474.
View Abstract

Stay Connected to Science

Navigate This Article