Comparative Genomics of the Sigatoka Disease Complex on Banana Suggests a Link between Parallel Evolutionary Changes in Pseudocercospora fijiensis and Pseudocercospora eumusae and Increased Virulence on the Banana Host

Abstract

The Sigatoka disease complex, caused by the closely-related Dothideomycete fungi Pseudocercospora musae (yellow sigatoka), Pseudocercospora eumusae (eumusae leaf spot), and Pseudocercospora fijiensis (black sigatoka), is currently the most devastating disease on banana worldwide. The three species emerged on bananas from a recent common ancestor and show clear differences in virulence, with P. eumusae and P. fijiensis considered the most aggressive. In order to understand the genomic modifications associated with shifts in the species virulence spectra after speciation, and to identify their pathogenic core that can be exploited in disease management programs, we have sequenced and analyzed the genomes of P. eumusae and P. musae and compared them with the available genome sequence of P. fijiensis. Comparative analysis of genome architectures revealed significant differences in genome size, mainly due to different rates of LTR retrotransposon proliferation. Still, gene counts remained relatively equal and in the range of other Dothideomycetes. Phylogenetic reconstruction based on a set of 46 conserved single-copy genes strongly supported an earlier evolutionary radiation of P. fijiensis from P. musae and P. eumusae. However, pairwise analyses of gene content indicated that the more virulent P. eumusae and P. fijiensis share complementary patterns of expansions and contractions in core gene families related to metabolism and enzymatic degradation of plant cell walls, suggesting that the evolution of virulence in these two pathogens has, to some extent, been facilitated by convergent changes in metabolic pathways associated with nutrient acquisition and assimilation. In spite of their common ancestry and shared host-specificity, the three species retain fairly dissimilar repertoires of effector proteins, suggesting that they likely evolved different strategies for manipulating the host immune system. Finally, 234 gene families, including seven putative effectors, were exclusively present in the three Sigatoka species, and could thus be related to adaptation to the banana host.

Fig 1. Disease symptoms caused on banana by the three species that constitute the Sigatoka disease complex.

(A–C) Leaf symptoms and conidia of Pseudocercospora eumusae. (D–F) Leaf symptoms and conidia of Pseudocercospora fijiensis. (G–I) Leaf symptoms and conidia of Pseudocercospora musae. Scale bars = 10 μm. (leaf photo credits Profs. A. Viljoen and G. Kema).

Fig 2. Genome size and composition in Pseudocercospora musae, Pseudocercospora eumusae, and Pseudocercospora fijiensis.

(A) Overall genome composition and repeat content in P. musae, P. eumusae, and P. fijiensis. The size (Mb) and proportion (%) of the main components of the species’ genomes are indicated. Given their close evolutionary relationships, the species show considerable differences in genome size mainly due to differences in repeat content. (B) The distribution and composition of repeat elements in P. musae, P. eumusae, and P. fijiensis. The proportion (%) and size (Mb) of each individual class of repeat elements are indicated. The three species differ in their composition and proportion of the different classes or repeat elements.

Fig 3. Molecular phylogeny of the three species that constitute the Sigatoka disease complex and 16 other representative Dothideomycetous fungi.

The maximum likelihood (ML) tree was constructed based on a concatenated sequence alignment of 46 orthologous single-copy genes. Bootstrap values (%) are indicated next to corresponding branching nodes. Aspergillus nidulans (class of Eurotiomycetes) was used as an outgroup species for rooting the tree. The selected 16 representative Dothideomycete species that are included in the phylogeny fall into three major orders, i.e. Capnodiales (red), Hysteriales (blue), and Pleosporales (green). In the inferred topology P. musae, P. eumusae, and P. fijiensis are strongly clustered (bootstrap value of 100%) as a monophyletic clade within the Capnodiales, whereas P. eumusae is sister to P. musae (bootstrap value of 100%), suggesting an earlier split of P. fijiensis from the common ancestor of these two species.

Fig 4. Shared and species-specific gene families and genes in Pseudocercospora musae, Pseudocercospora eumusae, and Pseudocercospora fijiensis.

(A) Venn diagram showing the total number of species-specific genes and shared gene families among the three species, as determined by reciprocal BlastP best hit (e-value: 1e-5) analysis implemented in OrthoMCL. A larger number of species-specific genes are found in P. fijiensis, whereas more gene families are shared between P. eumusae and P. fijiensis as compared to P. eumusae and P. musae, or P. musae and P. fijiensis. (B) The Venn diagram is expanded to include a broader comparison of the three species gene content against the NCBI nr database and the JGI fungal genome database (BlastP e-value: 1e-5, alignment coverage > 50%). In both Venn diagrams, the number of genes from each species included within the pool of shared gene families is indicated at every intersection.

Fig 5. Distribution of KOG annotation profiles in core gene families with or without copy number variation (CNV) among Pseudocercospora musae, Pseudocercospora eumusae, and Pseudocercospora fijiensis.

(A) Plotted in the different segments of the stacked bars is the percent of KOG terms assigned to core gene families (column 1), core gene families without (w/o) copy number variation (CNV) (column 2), and the core gene families with CNV (column 3), for each of the main functional categories of KOG (Cellular processes and signaling: blue; Information storage and processing: red; Metabolism: green; and Poorly characterized: purple). The number in the parenthesis of each segment in the columns refers to the number of KOG terms assigned to the gene families for each specific functional category of KOG. The first number in the X-axis label of each comparison refers to the total number of gene families with assigned KOG terms, whereas the second number refers to the total number of gene families in each comparison compartment. A high fraction (211, 50.2%) of the total number of 420 KOG terms that were collectivity assigned to gene families with CNV was ascribed to metabolism. (B) The number of gene families with CNV assigned to each subcategory of KOG. A high number of gene families (46) is associated with biosynthesis of secondary metabolites, transoport and catabolism, as well as carbohydrate transport and metabolism (42), lipid transport and metabolism (39), and amino acid transport and metabolism (38). Note that because some gene families receiving KOG annotations could be assigned to more than one functional categories of KOG, the number of KOG terms in this case is equivalent to the number of gene families.

Fig 6. Hierarchical clustering of Pseudocercospora musae, Pseudocercospora eumusae, and Pseudocercospora fijiensis based on copy number changes in different groups of KOG gene families.

Hierarchical clustering of the species based on (A) the KOG distribution profile (i.e. the number of genes assigned to each category of KOG) of their entire proteomes, (B) the KOG distribution profiles of the 575 core gene families with copy number variation (CNV), and (C) a subset of 190 core gene families with CNV that are predicted to be involved in metabolism based on KOG assignments. The reliability of the clustering patterns was assessed by bootstrap tests (1000 replicates) and obtained bootstrap values are indicated next to their corresponding branching nodes. While clustering of the species based on the KOG distribution profiles of the entire proteomes follows a pattern that is respective of the their phylogenetic relations (Fig 2), clustering of the species based on the KOG profiles of core gene families with CNV or their subset of gene families involved in metabolism, indicates a swapped topology in which P. eumusae is clustered together with P. fijiensis suggesting that these two species share a more similar pattern of gene family expansions and contractions.

Fig 7. Hierarchical clustering of Pseudocercospora musae, Pseudocercospora eumusae, Pseudocercospora fijiensis, and 16 other representative Dothideomycete fungi with different nutritional lifestyles, based on copy number changes in carbohydrate-active enzyme (CAZyme) families or the subset of plant cell wall degrading enzymes (PCWDEs).

The selected 16 representative Dothideomycete species that are included in the analysis fall into three major orders: Capnodiales (red), Hysteriales (blue), and Pleosporales (green). The nutritional lifestyle of each species is indicated by a colored dot next to each species name: biotrophs (blue), hemi-biotrophs (green), necrotrophs (yellow), saprophytes (red). (A) Hierarchical clustering of the species based on their total CAzyme distribution profile (i.e. the number of genes assigned to each CAZyme family) (B) Hierarchical clustering of the species based on their distribution profile for CAZyme families related to plant cell wall degradation. In both cases, clustering supported a swapped topology in which P. eumusae is clustered together with P. fijiensis, suggesting that these two species share a more similar pattern of gene family expansions and contractions in CAZymes and PCWDEs in particular.