Genome Expansion and Gene Loss in Powdery Mildew Fungi Reveal Tradeoffs in Extreme Parasitism

See allHide authors and affiliations

Science  10 Dec 2010:
Vol. 330, Issue 6010, pp. 1543-1546
DOI: 10.1126/science.1194573


Powdery mildews are phytopathogens whose growth and reproduction are entirely dependent on living plant cells. The molecular basis of this life-style, obligate biotrophy, remains unknown. We present the genome analysis of barley powdery mildew, Blumeria graminis f.sp. hordei (Blumeria), as well as a comparison with the analysis of two powdery mildews pathogenic on dicotyledonous plants. These genomes display massive retrotransposon proliferation, genome-size expansion, and gene losses. The missing genes encode enzymes of primary and secondary metabolism, carbohydrate-active enzymes, and transporters, probably reflecting their redundancy in an exclusively biotrophic life-style. Among the 248 candidate effectors of pathogenesis identified in the Blumeria genome, very few (less than 10) define a core set conserved in all three mildews, suggesting that most effectors represent species-specific adaptations.

Filamentous eukaryotes such as fungi and oomycetes (stramenopiles) are responsible for many serious plant diseases. Among these pathogens is a group of taxonomically diverse species, collectively termed obligate biotrophs, which only grow and reproduce on living plants. These microorganisms cause rusts, as well as downy and powdery mildews, and form dedicated invasive infection structures (haustoria) for nutrient uptake. Obligate biotrophs are found in two kingdoms (Stramenopila and Fungi) and in both major fungal phyla (Ascomycota and Basidiomycota), indicating that biotrophy is the result of convergent evolution. The ascomycete powdery mildews infect ~10,000 angiosperm species, including many important crops (1). They form morphologically complex structures during asexual pathogenesis and produce fruiting bodies (cleistothecia), which develop after sexual reproduction (Fig. 1 and fig. S1).

Fig. 1

Key developmental stages of powdery mildews and comparative ascomycete genome sizes. (A) Conidium 10 hours post-inoculation (hpi) showing the primary (arrowhead) and the appressorial germ tube (asterisk). Scale bar, 20 μm. (B) Appressorium (14 hpi) with penetration peg (p) and epidermal plant cell wall (cw). Scale bar, 1 μm. (C) Purified haustorium and host perihaustorial membrane (arrowhead). Scale bar, 10 μm. (D) Colony on barley, 4 days post-inoculation: conidiophores (c) and foot cells (fc). Scale bar, 100 μm. (E) Phylogeny of selected ascomycetes and their genome sizes (in megabases). Blumeria, G. orontii, and E. pisi are shown in red. The median genome sizes of the hemiascomycetes (blue vertical line, 12.3 Mb) and euascomycetes (green vertical line, 36.7 Mb) are also shown.

We sequenced the haploid Blumeria genome with the use of Sanger protocols and second-generation methods (table S1) (2). We assembled the sequence reads with a combination of the cortex and CABOG (Celera assembler with the best overlap graph) (3) assemblers into 15,111 contigs (L50: 18,024 bases; L50 is the length of the smallest N50 contig, where N50 is the minimum number of contigs required to represent 50% of the genome) on 6898 supercontig scaffolds (L50: 2,209,085 bases). The overall assembly size is 119,213,040 nucleotides (table S1). We estimate that the actual genome size is ~120 Mb, corresponding to 140-fold coverage of the Blumeria genome. We additionally generated draft genome assemblies (~eightfold coverage each) of two other powdery mildew species, Erysiphe pisi [pathogenic on pea (Pisum sativum)] and Golovinomyces orontii (pathogenic on Arabidopsis thaliana). Together with Blumeria, these species represent three of the five major tribes of the order Erysiphales, which diverged ~70 million years ago (4). We calculated that the genome sizes of the latter two species are ~151 and ~160 Mb, respectively (table S1). Thus, the genome size of each of the mildews is more than four times larger than the median of other ascomycetes (Fig. 1E). We first annotated the Blumeria genome using ab initio gene finders followed by extensive manual curation (table S2). The actual number of curated genes is 5854, which is at the lower end of the range of fungal genomes.

The comparatively low gene number and the inability of the parasite to grow in vitro suggest that the mildew genomes may lack genes typically present in autotrophic ascomycetes. We systematically searched for genes absent in the mildews but present in baker’s yeast (Saccharomyces cerevisiae) and the phytopathogens Colletotrichum higginsianum, Magnaporthe oryzae, and Sclerotinia sclerotiorum (Fig. 2A). We identified 90 yeast genes by this procedure and 9 additional common ascomycete genes by manual inspection that are missing in the genome assemblies of all three mildews [hereafter referred to as missing ascomycete core genes (MACGs)]. It is unlikely that these gene sets were missed as a result of incomplete genome coverage because (i) the 140× Blumeria assembly encompasses >99% of the conserved gene space (table S1), and (ii) these genes are missing in all three assemblies. The MACGs represent a diverse set of metabolic and regulatory proteins, affecting multiple processes and pathways (for example, thiamine biosynthesis), and a considerable subset of MACGs (57 to 77%) also seems to be absent in other obligate biotrophic phytopathogens (fig. S2 and table S3).

Fig. 2

Gene losses in powdery mildews. (A) Number of missing S. cerevisiae proteins in one fungus compared with the three others as a function of TBLASTN e-values. (B) Number of genes devoted to secondary metabolism and genes encoding cellulose- or hemicellulose-degrading enzymes in Blumeria (left) and M. oryzae (right). DMATS, dimethylallyl diphosphate tryptophan synthase; GH1, glycosyl hydrolase.

The existence of MACGs raises the possibility that their expression may be detrimental to biotrophy. To test this, we determined expression of MACG homologs in C. higginsianum, a phytopathogenic fungus that first employs a biotrophic growth mode and later switches to necrotrophic pathogenesis, involving host cell killing. Analysis of the C. higginsianum transcriptome revealed that most of the MACG homologs that we tested (26 out of 32) are expressed during the biotrophic stage (fig. S3). Their expression is, therefore, unlikely to be detrimental to biotrophic growth.

Although the vast majority of genes encoding enzymes of primary metabolism are retained, notable exceptions include anaerobic fermentation, biosynthesis of glycerol from glycolytic intermediates, and inorganic nitrogen (nitrate) assimilation. These deficiencies are consistent with an exclusively aerobic, parasitic life-style on aerial plant organs, the production of solutes for the generation of osmotic pressure during plant cell wall penetration from triacyl glycerol breakdown, and the assimilation of organic nitrogen in the form of host-derived amino acids.

Filamentous fungi generally produce an array of secondary metabolites, some of which are involved in pathogenesis (5). Key enzymes that catalyze their biosynthesis include polyketide synthases (PKSs), modular nonribosomal peptide synthetases (NRPSs), terpene cyclases, and dimethylallyl diphosphate tryptophan synthases (6). Blumeria encodes only two such proteins (one PKS and one NRPS), the lowest number known in fungi (Fig. 2B and fig. S4). We hypothesize that Blumeria synthesizes only one iron siderophore and one simple polyketide, possibly the pigment observed on cleistothecia (fig. S1, G and H). Similar trends are observed in other biotrophs, such as the basidiomycete Ustilago maydis and the plant symbiotic fungus Tuber melanosporum. Therefore, it appears that biotrophy is associated with a convergent loss of secondary metabolic enzymes. We also noted a marked reduction in genes encoding specific subfamilies of transporters (fig. S5), which typically function in secretion of toxins into the host and extrusion of host defense compounds in necrotrophic fungi (7).

Unlike other known plant pathogenic fungi, Blumeria has an extremely reduced set of carbohydrate active enzymes devoted to plant cell wall depolymerization (Fig. 2B and fig. S6) (8). We found no canonical cellulose-, xylan-, or pectin-degrading enzymes. Other biotrophic phytopathogens, such as U. maydis and Puccinia graminis, also possess reduced enzyme systems for degradation of the plant cell wall, but both species have predicted cellulases and xylanases. An example of structural proteins lacking in the mildews are the hydrophobins, a class of cell wall proteins that are typically present in fungi (9).

We found a massive proliferation of transposable elements (TEs) (table S4). In Blumeria, where TEs account for 64% of the genome size, the most abundant families comprise non–long terminal repeat (LTR) retrotransposons lacking LTRs (fig. S7). TEs were evenly distributed throughout the Blumeria genome, with no evidence of clustering of particular TEs (fig. S8). Protein-coding genes are typically in small clusters (2 to 10 genes) interspersed between extended stretches of TEs. In all three powdery mildew genomes, genes required for repeat-induced point mutations (RIPs) are absent, whereas all components known to be necessary for mitotic and meiotic silencing are present (table S5). Thus, dysfunctionality of the RIP pathway has probably contributed to genome-size inflation, and extensive retrotransposition (rather than gradual pseudogenization) may account for the observed gene losses and reshuffling. An example of the latter is the otherwise-conserved mating type (MAT) locus (10). In all other ascomycetes, MAT genes are flanked by a conserved cluster of functionally unrelated genes. Although the microsynteny of these genes is retained in the Blumeria genome (fig. S9), MAT was found on a different supercontig, indicating physical separation.

In addition to >1350 paralog copies of the previously described atypical avirulence genes AVRk1 and AVRa10 (11), we predicted 248 Blumeria proteins with a signal peptide (SP) but lacking any transmembrane domain and BLAST (Basic Local Alignment Search Tool) hit outside the mildews, thus representing candidates for secreted effector proteins (CSEPs) (12). The CSEPs have distinctive features (table S6) and show great sequence diversity with few members grouping in small families (Fig. 3A). We noted no obvious clustering of CSEPs within the Blumeria contigs. Approximately 80% harbor a recently identified N-terminal tripeptide motif, termed “YxC,” (13), that typically occurs within the first 30 amino acids after the predicted SP cleavage site. Searches in the E. pisi and G. orontii genomes revealed that the vast majority of CSEPs are confined to Blumeria (Fig. 3A and table S6). Thus, powdery mildew genomes preferentially harbor species- and/or tribe-specific innovations, which possibly evolved in the context of cospeciation with their plant hosts (11). Upon comparison of global gene expression in haustoria (Fig. 1C) and epiphytic structures (Fig. 1D), we observed preferential expression of the majority of the CSEPs (79%) in haustoria (Fig. 3B), suggesting they have specific functions in biotrophic pathogenesis (14).

Fig. 3

Diversity and expression profiling of candidate secreted effector genes. (A) Circular neighbor-joining phylogenetic tree of the 248 CSEPs. Scale, amino acid substitutions per site; green, CSEPs harboring the YxC motif; blue, CSEPs conserved in E. pisi and/or G. orontii; red, CSEPs predominantly expressed in haustoria. (B) Global gene expression in haustoria (H) versus epiphytic structures (E). Relative abundance of each gene plotted versus the p-value as a measure of the statistical significance. White circles, all gene models; black, no SP and no TM domain; yellow, TM only; red, CSEPs; pink, genes with BLASTP hits in the National Center for Biotechnology Information nr database and SP only; light blue, both SP and TM domains.

We detected common genomic hallmarks in the powdery mildews associated with obligate biotrophy. These include gene losses and extensive gene reshuffling correlated with expansion in (retro-)transposon number and genome size. Together, these hallmarks may represent a tradeoff between advantages of increased genetic variation independent of sexual recombination and irreversible deletion of genes dispensable for biotrophy. Hence, their evolution provides a notable example of Dollo’s law of evolutionary irreversibility (15). This may explain why powdery mildews and possibly other biotrophic parasites became obligate.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S9

Tables S1 to S6


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank B. Thomas, T. Carver, and D. Grunewald for providing scanning electron micrographs. This work was supported by funds from the Biotechnology and Biological Sciences Research Council [grant BB/E0009831/1], INRA, the BioExploit European Union Framework 6 project, Deutsche Forschungsgemeinschaft (priority program SPP1212), the Leverhulme Trust, and the Max Planck Society. GenBank project identification numbers are as follows: Blumeria, 28821; E. pisi, 50315; and G. orontii, 50317.
View Abstract

Navigate This Article