Introduction

The range of hosts that pathogens can infect is determined by genetic and environmental factors. This host range is an important factor in assessing the dynamics of disease epidemics. Specialists parasitize one or few hosts, such as the cereal powdery mildews infecting only cereals [1] while generalists live on a range of diverse hosts. Natural selection by host populations and environmental factors drives frequent host switches and variations in pathogen–host range. Specialization occurs when a pathogen adapts to a specific host and enters a co-evolutionary arms race with it. In some instances, adaptation following host switching and host jumps involves the ability to efficiently colonize new hosts while retaining the ability to colonize the original host lineage, resulting in host range expansions [2, 3].

Adaptation to the gene-for-gene type of plant resistance is a paradigmatic example of co-evolutionary arms race [4]. Plant resistance genes typically recognize specific pathogenic proteins called effectors and mount a resistance reaction upon perception. Adapted pathogens evolved to avoid recognition by modification or loss of the respective effector [5]. This involves rapid adaptation, for example by selective sweeps [6] leaving characteristic patterns of variation in the genome of plant pathogens [7, 8]. In the potato late blight pathogen Phytophthora infestans, the resulting genetic variation is notably responsible for a tradeoff in effector activity on targets from different hosts [9] and distinctive two-speed-genome architecture [10, 11]. In addition, balancing selection allows polymorphisms to persist in the gene pool and increases the standing genetic diversity in populations [12]. Arms race models generally assume an isolated pathogen co-evolving with one host via pairwise selection. However, pathogen genomes often evolve in response to selection caused by more than one host under diffuse co-evolution [13, 14]. Genomic signatures of diffuse selection and molecular adaptations associated with interaction with multiple hosts are largely unresolved [14].

Theories of evolutionary transitions suggest that genetic accommodation of pathogens to new hosts could entail general-purpose molecular bases supporting the colonization of any host (true generalist) or multiple modules dedicated to the colonization of specific hosts (polyspecialist) [3, 15]. Core effectors that act on conserved plant targets could serve as general-purpose molecules contributing to the accommodation of new hosts [2]. For example, the plant peroxidase inactivating effector PEP1 is conserved among the Ustilaginaceae and functional even in distantly related non-host plant species [16]. Comparative genomic studies have highlighted the role of the expansion of gene families related to secondary metabolism and detoxification in host range expansion in insect [17] and fungal [18] plant pathogens. A recent comparative study reported the transition from specialized one-speed genomes toward adaptive two-speed genomes correlated with increased host range in ergot fungi [19]. However, P. infestans and the head blight pathogen Fusarium graminearum exhibit typical two-speed-genome architecture and high degrees of host specialization, indicating that genome transitions cannot be unambiguously associated to directional host range variation [10, 11, 20]. The analysis of seven fungi from the Metarhizium genus of entomopathogens showed an expansion in genes encoding G protein-coupled receptors, proteases, transporters, enzymes for detoxification, and secondary metabolite biosynthesis, coinciding with increased host range [21]. In this lineage, horizontal gene transfers contributed to host range expansion [22]. By contrast, genome size was inversely correlated with host range in Helicoverpa butterflies [23]. The genome of the generalist aphid Myzus persicae has a gene count half that of Acyrthosyphon pisum, which is specialized on pea. Instead, M. persicae colonizes diverse host plants through rapid transcriptional induction of specific gene clusters [24]. These studies support the polyspecialist model of host range expansion in which pathogen genomes harbor several specialized gene modules, resulting from gene family expansion or differential gene expression.

The white mold fungus Sclerotinia sclerotiorum is a typical generalist plant pathogen reported to infect more than 400 plant species [25]. Its genome lacks signatures of selective sweeps [26] and two-speed architecture [27]. Host colonization by S. sclerotiorum is supported by division of labor enabling cooperation between cells of invasive hyphae [28]. In addition, the S. sclerotiorum genome shows signatures of adaptive translation, a selective process shaping the optimization of codon usage to increase protein synthesis efficiency. Codon optimization is particularly clear in genes expressed during plant colonization and encoding predicted secreted proteins [29]. Division of labor and codon optimization increase S. sclerotiorum fitness independently of the host genotype and therefore constitute genomic signatures of a true generalist. In addition, a study of a few effector candidate genes suggested the existence of plant host-specific expression patterns [30], and various small RNAs are differentially expressed on Arabidopsis thaliana and common bean (Phaseolus vulgaris) [31]. Furthermore, various isoforms originating from host-specific alternative splicing add additional plasticity to the transcriptomic profile of S. sclerotiorum [32] and suggest polyspecialist adaptations in this species. Nevertheless, the extent to which S. sclerotiorum genome harbors signatures of adaptation to a polyspecialist lifestyle has not been fully elucidated.

Here, we tested whether S. sclerotiorum strain 1980 activates distinct gene sets for the infection of diverse host species. Using global gene expression data obtained during the infection of plants covering six botanical families [28, 32, 33] we reveal core and flexible host-specific transcriptional programs. Genes activated specifically during the colonization of plants from the Brassicaceae family are largely conserved in the newly sequenced genome of the sister species Sclerotinia trifoliorum strain SwB9, unable to infect Brassicaceae plants. We show that lack of transcriptional response to camalexin in SwB9 coincides with increased sensitivity of several S. trifoliorum strains to this Brassicaceae defense compound. Interspecific comparison of promoter sequences suggests that regulatory variation may associate with the genetic accommodation of Brassicaceae in the Sclerotinia host range. Our work associates adaptive plasticity of a broad host range pathogen with specific responses to different host plants and exemplifies the co-existence of signatures for generalism and polyspecialism in the genome of a plant pathogen.

Materials and methods

Biological material

The fungal isolates S. sclerotiorum isolate 1980 [25] and S. trifoliorum SwB9 [34] were used in this study. These strains were chosen for their strong, but not extreme, aggressiveness on their respective hosts [34, 35] and for originating both from hosts in the Fabaceae family. Camalexin sensitivity was also tested for S. sclerotiorum strains Fr.B5 [34], CU6.1, MB21, MB52, S55 [26], Blo.P144, and Blo.P154, S. Trifolium strains Be.B9, Li.A6 and Sw.B8 [34], Sclerotinia minor CBS 112.17, Sclerotinia nivalis MAFF 21347, and Sclerotinia kitajimana MAFF 410428. The fungi were cultivated on potato dextrose agar (PDA) at 23 °C or stored on PDA at 4 °C. A. thaliana accession Col-0 and the T-DNA insertion lines cyp79b2 cyp79b3 [36] and pad3-1 [37] were grown in Jiffy pots under controlled conditions at 23 °C, with a 9-h light period at intensity of 120 µmol m−2 s−1 for up to 5 weeks.

RNA sequencing and gene expression analyses

RNA was collected in triplicates as described in [28] and [33], and from fungi cultivated on potato dextrose agar (Fluka) with DMSO or camalexin (125 µM for S. sclerotiorum, 25 µM for S. trifoliorum). Quality and concentrations of RNA were assessed with Agilent bioanalyzer nanochips. RNA sequencing (RNA-seq) was performed by Fasteris (Switzerland, Plan-les-Ouates) to produce Illumina reads (125 bp, paired-end) on a HiSeq2500 sequencer. Quality and adapter trimming of reads and reads mapping were performed as in [38] against the S. sclerotiorum isolate 1980 reference genome [27]. FPKM (fragments per kilobase of transcript per million mapped reads) tables were generated using the Cufflinks function cuffnorm with -compatible-hits-norm -library-norm-method classic-fpkm [39] (Dataset S1: Table S1). Differential expression analysis was run on 7423 protein-coding genes under a limma-edgeR pipeline [40] with cut-offs of p ≤ 0.01 and log2 fold change (LFC) ≤−1 (downregulated) or ≥1 (upregulated) using S. sclerotiorum gene expression on PDA as a reference (Dataset S1: Table S2). R v.3.5.1 [41] was used for statistical analysis and generation of plots. Codes for data analysis are deposited at https://github.com/stefankusch/sclerotinia_2020. GO and PFAM annotations (Dataset S1: Table S3, Table S4) and enrichment analyses were performed as described in [33].

S. trifoliorum genome assembly and comparative analysis

High molecular weight DNA isolation was performed as in [38]. Library preparation and sequencing were done at the GeT-PlaGe core facility, INRAE Toulouse, France, as in [38] with the following modifications: For one flow cell, 5 µg of purified DNA was sheared at 8 kb using the megaruptor1 system (Diagenode), followed by a DNA damage repair step on 2 µg of sample. Then an END-repair, dA-tailing of double-stranded DNA fragments, and adapters ligation were performed on the library. The library was loaded onto one R9.4.1 flow cell and was sequenced on a GridION instrument at 0.05 pmol within 48 h. The Canu v1.6 [42] assembly yielded 48 contigs between 3,548,298 and 6,307 bp and a total genomic length of 40,161,953 bp. We did four cycles of polishing with Pilon [43] and then used Blobtools v1.0 [44] to identify contigs with Sclerotiniaceae identity, as described before [38] (Fig. S1). The final genome assembly (Table 1) was subjected to repeat masking (RepeatMasker v4.0.7 [45], RepBase-20170127) prior to ab initio gene annotation with BRAKER2 [46] where we included all RNA-seq data of S. trifoliorum SwB9 produced in this study, i.e., cultivated in vitro on PDA (1×), PDA with DMSO (3×), and with 25 µM camalexin (3×), and from infection of P. vulgaris (1×). All gene models were manually curated via Web Apollo [47]. The genomes of S. sclerotiorum Ss1980 [27] and S. trifoliorum SwB9 were compared by synteny using MUMmer3; [48] synteny plotting was performed with genoPlotR [49]. The proteomes of S. sclerotiorum Ss1980 [27], Myriosclerotinia sulcatula MySu01 [38], Botrytis cinerea B05.10 [50], and S. trifoliorum SwB9 were compared via OrthoFinder [51].

Table 1 S. trifoliorum SwB9 genome assembly statistics.
Full size table

Detection of cis-elements

The 1000-bp upstream sequences of all genes of S. sclerotiorum Ss1980 [27] and S. trifoliorum SwB9 were extracted using bedtools v2.25.0 [52]. The MEME-Suite 5.1.1. at http://meme-suite.org [53, 54] was used for motif discovery and enrichment (MEME, DREME), motif scanning (FIMO), and motif comparison (TOMTOM) against the Saccharomyces cerevisiae YEASTRACT_20130918.meme database. The S. cerevisiae motif-binding proteins were queried against the S. sclerotiorum Ss1980 proteome via BLASTP to identify the potential motif-binding orthologues DNA-binding proteins.

Results

Host-specific transcriptome reprogramming in S. sclerotiorum

To investigate transcriptional reprogramming in S. sclerotiorum during the colonization of hosts from different botanical families, we performed RNA-seq analysis during infection of A. thaliana, Solanum lycopersicum, Helianthus annuus, P. vulgaris, Ricinus communis and Beta vulgaris (Fig. 1A). To control for variations in the kinetics of pathogen colonization on different hosts, samples were collected at similar infection stages by macro-dissection of disease lesion edge [28]. We performed differential expression analysis on 7524 expressed genes using mycelium grown in vitro as reference. We identified 1712 differentially expressed genes (DEGs, 1120 genes upregulated and 592 genes downregulated) in lesion edge across six hosts. The number of upregulated genes ranged from 110 on H. annuus to 777 on B. vulgaris (7.1-fold variation) and the number of downregulated genes ranged from 16 on S. lycopersicum to 409 on R. communis (25.6-fold variation) (Fig. 1A). Hierarchical clustering and principal component analysis showed clear separation of S. sclerotiorum gene expression according to the infected host (Fig. 1B, Fig. S2, Dataset S1: Tables S5 and S6). In colony edge, 53 DEGs (4.7%) were upregulated during the colonization of all host plants, 483 DEGs (43.1%) where upregulated on at least two host plants, and 584 DEGs (52.1%) showed specific upregulation on one host (Fig. 1C). The number of genes upregulated on one host only represented 0% on S. lycopersicum, 4.5% on P. vulgaris, 14.7% on A. thaliana, 15.5% on H. annuus, 32.3% on R. communis and 39.6% on B. vulgaris. These results indicate that the colonization of some hosts relies on core virulence genes while other hosts trigger the activation of host-specific fungal virulence programs.

Fig. 1: Differential regulation of S. sclerotiorum transcriptome during the colonization of plants from six botanical families.
figure 1

A Number of S. sclerotiorum genes expressed differentially (DEGs) on each host compared to in vitro-grown colonies. Bubbles are sized proportionally to the number of DEGs in each treatment, labels showing the corresponding number of genes. The upper half of bubbles correspond to genes upregulated, the lower half to genes downregulated. B Hierarchical clustering of RNA-seq samples based on the expression of DEGs. Numbers in branch labels correspond to biological replicates. C Distribution of DEGs according to host species. For each sector, the upper value (Δ) corresponds to upregulated genes, the lower value (∇) to downregulated genes. The central dark gray hexagon shows DEGs detected on all six hosts, the light gray hexagon shows DEGs detected on 2 to 5 hosts. D Hierarchical clustering of log2 fold change of expression for 59 expressed genes encoding cytochrome p450. Eight hierarchical clusters labeled a-h were delimited. Ath, Arabidopsis thaliana; Bvu, Beta vulgaris; Han, Helianthus annuus; Pvu, Phaseolus vulgaris; Rco, Ricinus communis; Sly, Solanum lycopersicum.

Full size image

To document the functional diversity of S. sclerotiorum genes differentially expressed during host colonization, we analyzed GO and PFAM annotation enrichment with genes upregulated in planta. We found 11 GOs and 94 PFAMs significantly enriched with upregulated genes (chi-squared adjusted p val < 0.01) during the colonization of at least one plant (Dataset S1: Table S7 and  S8). Thirteen of these PFAMs were enriched in genes upregulated during the colonization of five or six host species, including galactosidase (PF10435, PF13363, PF13364, PF16499), glycosyl hydrolase (GH) (PF00150, PF01301, PF00295), sugar transporter (PF00083), cytochrome p450 (PF00067) and domain of unknown function DUF4965 (PF16335) domains. Forty PFAMs were enriched in genes upregulated during the colonization of a single host species (Fig. S3). Next, we examined the differential expression pattern of the largest gene families enriched in upregulated genes using hierarchical clustering of LFC values. We identified 59 expressed genes harboring a cytochrome p450 domain into eight hierarchical clusters (a–h, Fig. 1D). The colonization of each host lead to a specific signature of p450 genes upregulated in S. sclerotiorum. We identified 147 expressed genes harboring a GH domain that classified into 11 hierarchical clusters (Fig. S4). The colonization of each host upregulated specific sets of GH genes. The 1120 upregulated genes included 246 secretome genes (21.9%), coding for putative secreted proteins. Secretome genes represented 8.5% of the 7423 expressed genes, 18.3% of genes upregulated on one host, 23.2% of genes upregulated on two, three, or four hosts, and 35.5% of genes upregulated on five or six hosts. This enrichment is consistent with a prominent role of fungal secreted proteins in the interaction with host plants, and indicate that secreted proteins contribute to a larger part of the S. sclerotiorum core infection program than the host-specific infection programs.

Arabidopsis-specific induced genes are conserved in the genome of the nonpathogenic S. trifoliorum

S. trifoliorum is closely related to S. sclerotiorum but has a host range restricted to plants in the Asterids and Fabids families [55]. Unlike S. sclerotiorum, S. trifoliorum colonizes A. thaliana very poorly to not at all. To gain insights into the evolution of A. thaliana-specific upregulated genes in the Sclerotinia lineage, we performed genome sequencing of S. trifoliorum isolate SwB9 using Illumina short-read and nanopore long-read data. We assessed completeness of the genome with BUSCO [56] and found 97.4% of highly conserved ascomycete genes present, suggesting near-complete gene space coverage. According to synteny analysis 67.2% of the S. trifoliorum genome aligned to 68.8% of the S. sclerotiorum genome, with a number of large-scale inversions and an apparent chromosome arm exchange (Fig. 2A, B). Overall, we generated a near-chromosome assembly for S. trifoliorum SwB9 of sufficient quality for gene space comparison with S. sclerotiorum. We predicted 10,626 unique gene models in the genome of S. trifoliorum SwB9. The S. trifoliorum proteome comprises 2,020 proteins with putative transmembrane domains and 705 proteins contained predicted secretion signal peptides, 73 of which are possible effector candidates according to an EffectorP 2.0 search [57]. We found 9155 orthogroups (OGs) shared between the two Sclerotinia species containing 10,085 (S. sclerotiorum) and 9922 (S. trifoliorum) genes, respectively (Fig. 2C, Dataset S1: Table S9). 685 S. sclerotiorum genes had no ortholog in S. trifoliorum, including 249 S. sclerotiorum genes from 199 OGs that did not contain S. trifoliorum genes. Out of 306 S. sclerotiorum genes induced at the edge of colonies on A. thaliana, 300 (98%) had orthologs in S. trifoliorum. Out of 45 S. sclerotiorum genes upregulated specifically on A. thaliana, 42 (93.3%) had orthologs in S. trifoliorum (Dataset S1: Table S10). Twenty-seven of the 457 (5.9%) DEGs on the common host P. vulgaris were absent in S. trifoliorum, while all 28 P. vulgaris-specific DEGs had S. trifoliorum orthologues. In total, 1625 of the 1712 DEGs in S. sclerotiorum had orthologues in S. trifoliorum SwB9 (95.0%). Therefore, expansion of the host range of Sclerotiniaceae fungi to A. thaliana largely relied on genes acquired prior to the divergence between S. sclerotiorum and S. trifoliorum.

Fig. 2: Synteny and orthology relationships between S. sclerotiorum and S. trifoliorum genomes.
figure 2

A Overall synteny between S. sclerotiorum chromosomes (red) and S. trifoliorum contigs (gray). Colored ribbons connect syntenic regions across genomes. Chromosomes and contigs shown in B are labeled with bold fonts. B Synteny of S. sclerotiorum chromosomes 6 and 12 against the S. trifoliorum assembled contigs. C Venn diagram summarizing the results from the Orthofinder analysis between the proteomes of S. sclerotiorum 1980, S. trifoliorum SwB9, Myriosclerotinia sulcatula MySu01, and Botrytis cinerea B05.10. Values in bold correspond to number of orthogroups, values in italics correspond to number of genes.

Full size image

S. trifoliorum is sensitive and not responsive to phytoalexins

Plants in the order Brassicales like A. thaliana produce tryptophan-derived defensive metabolites such as indolic glucosinolates (iGLs), the indole alkaloid camalexin, and the indole phytoalexin brassinin. The infection of plants from the Brassicales by some pathogens involves metabolizing these plant defense compounds into non-toxic derivatives [58, 59]. To test whether the toxicity of tryptophan-derived plant defense metabolites could explain the inability of S. trifoliorum to colonize Brassicales, we compared the sensitivity of S. sclerotiorum and S. trifoliorum to five major defense tryptophan-derivatives produced by A. thaliana using an in vitro growth assay (Fig. 3A). Both S. sclerotiorum and S. trifoliorum tolerated similar concentrations of tryptophan, raphanusamic acid, indole-3-carboxylic acid and indole-3ylmethylamine (Figure S5). However, in contrast to S. sclerotiorum the growth of S. trifoliorum was completely inhibited by 125 µM camalexin and 250 µM brassinin. The four S. trifoliorum strains we tested were unable to grow on 125 µM camalexin, while all eight S. sclerotiorum strains tested grew without major defects (Fig. S5). The growth of strains of the closely related S. minor, S. nivalis, and S. kitajimana was also drastically impaired on 125 µM camalexin (Fig. S5). To verify that these compounds contribute to plant resistance to S. trifoliorum, we compared the colonization of wild type and cyp79b2/b3 A. thaliana plants, defective in iGLs biosynthesis [60] (Fig. 3B). Three days post inoculation (dpi), S. sclerotiorum had fully colonized 56% of Col-0 wild type leaves, while S. trifoliorum hardly grew out of the inoculation plug (0% of leaves fully colonized), consistent with the inability of S. trifoliorum to colonize A. thaliana. The cyp79b2/b3 mutant was more susceptible to S. sclerotiorum, harboring 89% of leaves fully colonized at 3 dpi. Remarkably, cyp79b2/b3 plants appeared susceptible to S. trifoliorum in this assay, as 67% of leaves were fully colonized at 3 dpi. These results indicate that sensitivity to tryptophan-derived defense metabolites contribute to the inability of S. trifoliorum to infect A. thaliana.

Fig. 3: S. trifoliorum is more sensitive than S. sclerotiorum to phytoalexins in vitro and in planta.
figure 3

A Phytoalexin tolerance plate assay. S. sclerotiorum 1980 and S. trifoliorum SwB9 were cultivated on potato dextrose agar (PDA) containing phytoalexins at different concentrations. The solvent used for each compound is indicated between brackets. Photos were taken after seven days; the experiment was conducted three times with similar results. Scale bar: 1 cm. B The A. thaliana wild type Col-0 and the indole glucosinolate and camalexin deficient mutant cyp79b2 cyp79b3 were infected with S. sclerotiorum 1980 and S. trifoliorum SwB9. Photos were taken three days after inoculation. Arrowheads indicate agar plugs with S. sclerotiorum (red) or S. trifoliorum (yellow). Scale bar: 1 cm. The bar chart indicates the proportion of inoculated leaves fully colonized by each fungus for n = 9 or 10 leaves. C FPKM expression values for differentially expressed genes during the colonization of A. thaliana (Ath, S. sclerotiorum only) and upon camalexin treatment (Cam., S. sclerotiorum and S. trifoliorum). Expression of orthologs of DEGs from the other Sclerotinia species are shown for comparison purposes. D Venn diagram illustrating the number of DEGs in S. sclerotiorum during the colonization A. thaliana, S. sclerotiorum growth on camalexin and orthologs of S. trifoliorum DEGs during growth on camalexin. The number between brackets corresponds to complete gene sets. E Relative expression at 72 h post inoculation for two S. sclerotiorum genes determined by quantitative reverse transcription PCR (Q RT-PCR) on A. thaliana wild type plants, cyp79b2/cyp79b3 and pad3 mutants. Values shown are for 6 independent biological replicates averaged over two technical replicates. DMSO dimethyl sulfoxide, PDA potato dextrose agar.

Full size image

This prompted us to explore the extent to which S. sclerotiorum and S. trifoliorum differ in their transcriptional response to camalexin. For this, we performed RNA sequencing of S. sclerotiorum and S. trifoliorum colonies grown on PDA plates with camalexin (Dataset S1: Table S11). We identified 323 genes upregulated in S. sclerotiorum on camalexin (Dataset S1: Table S12). Among those, 180 (55.7%) were also induced in at least one of the plant species (Dataset S1: Table S12) and 301 had orthologs in S. trifoliorum. Only 42 genes were upregulated in S. trifoliorum on camalexin (Dataset S1: Table S12), among which 40 had orthologs in S. sclerotiorum. Hierarchical clustering of expression for the 341 DEGs and their orthologs from the two species identified a cluster (b) of 70 upregulated during the colonization of A. thaliana and on camalexin in S. sclerotiorum (Fig. 3C, D). Only 3 orthologs of these genes in S. trifoliorum were induced on camalexin (Fig. 3D, Dataset S1: Table S12). We detected 23 genes encoding putative secreted proteins in this dataset of which 22 were conserved in S. trifoliorum SwB9. Nineteen of the respective proteins were predicted to be secreted in S. trifoliorum (Dataset S1: Table S13). One of these (Sscle05g046060, SCLTRI_001855) encoded an effector candidate according to EffectorP 2.0 analysis. The putative effector is 87 amino acids in length in both species, 12 of which were variable between the two species (Fig. S6). A BLASTP search against the nr database (E < 1e−10) revealed broad conservation of this effector in Sclerotiniaceae and in Fusarium sp., suggesting that Sscle05g046060 could represent a core effector with a conserved function in virulence against Brassicaceae.

We propose that the 70 S. sclerotiorum genes upregulated on camalexin and during the colonization of A. thaliana could respond to camalexin in planta. To confirm that camalexin and iGLs produced by A. thaliana are required to trigger the induction of S. sclerotiorum genes, we compared by quantitative RT-PCR the expression of seven S. sclerotiorum genes during the colonization of wild type, cyp79b2/b3, and pad3 plants, defective in camalexin biosynthesis (Fig. 3E, Fig. S7). At 72 hpi, the expression of Sscle07g055350 and Sscle15g106410 was not different during infection of A. thaliana wild type and mutant plants. The expression of Sscle02g011950 and Sscle02g022130 was strongly reduced during infection of cyp79b2/b3 but not pad3 as compared to wild type. The expression of Sscle04g037240, Sscle08g067130, and Sscle16g108230 was significantly reduced both during infection of cyp79b2/b3 and pad3 mutants as compared to wild type (Welch’s t test p value < 0.05). These results indicated that host-derived camalexin modulates the expression of S. sclerotiorum genes during infection but that S. trifoliorum does not significantly reprogram its transcriptome in response to this compound.

Cis-regulatory variation associates with the evolution of camalexin responsiveness in the Sclerotinia genus

We propose that transcriptome plasticity in host responsive genes contributed to host range variation in the Sclerotiniaceae. In particular, regulatory variation in the 70 S. sclerotiorum genes upregulated both on A. thaliana and on camalexin in vitro (Fig. 3, Dataset S1: Table S13) may have facilitated the colonization of plants from the Brassicaceae. To support this hypothesis, we first analyzed the conservation of these 70 genes in 670 species across the fungal kingdom (Figure S8). 91% of these genes were detected in more than 100 fungal species, suggesting that gene presence/absence polymorphism played a limited role in the evolution of responsiveness to plant-derived camalexin in S. sclerotiorum. Second, we analyzed the cis-elements in the 1000-bp upstream sequences of these genes. The WWCCCCRC motif was significantly enriched in these 70 S. sclerotiorum genes but not in the upstream sequences of the 69 orthologous S. trifoliorum genes (Fig. 4A). Using published yeast protein-DNA ChIP data we found five proteins of Saccharomyces cerevisiae, Mig1, Mig2, Mig3, Adr1, and Rsf2, that are capable of binding WWCCCCRC-like motifs. By homology searches and phylogenetic analyses, we identified potential orthologous proteins in S. sclerotiorum and S. trifoliorum (Fig. S9, File S1). Probable orthologs of yeast Mig1-3 (Multicopy Inhibitor of GAL gene expression) and Aspergillus CreA (Carbon catabolite repressor A) encoding a protein harboring a central zinc-finger H2C2 domain were Sscle01g002690 and Scltri_002966 (Fig. 4B). The expression of this gene was detectable in both fungi at FPKM > 100 and was slightly induced by camalexin (LFC 0.60) and during A. thaliana colonization (LFC 0.38) in S. sclerotiorum (Fig. 4C). Both Sscle01g002690 and Scltri_002966 contained two and four occurrences, respectively, of the WWCCCCRC motif in their upstream sequences. Probable orthologs of yeast ADR1 (Alcohol Dehydrogenase II synthesis Regulator) and RSF2 (Respiration factor 2) encoding a protein harboring a N-terminal zinc-finger H2C2 domain and a C-terminal fungal specific transcription factor domain PF04082 were Sscle07g055670 and Scltri_004270 (Fig. 4B). The expression of CreA (FPKM > 100) and ADR1 (FPKM > 20) was detectable in vitro in S. trifoliorum and S. sclerotiorum and in planta in S. sclerotiorum. They are therefore prime candidates for mediating large-scale transcriptome reprogramming in response to camalexin in the Sclerotinia lineage. These results suggest that cis-regulatory variation in targets of SsCreA and SsADR1 contributed to the evolution of transcriptional responsiveness to camalexin in S. sclerotiorum (Fig. 4C).

Fig. 4: Cis-regulatory variation in S. sclerotiorum genes induced by camalexin and during the colonization of A. thaliana.
figure 4

A Top: sequence logo of the WWCCCCRC cis-regulatory element enriched in S. sclerotiorum genes induced by camalexin and A. thaliana infection by comparison to their S. trifoliorum orthologs. Bottom: distribution of the WWCCCCRC element in the 1 kbp upstream sequence of S. sclerotiorum genes induced by camalexin and A. thaliana infection and their S. trifoliorum orthologs. B Domain structure of S. sclerotiorum orthologs of CreA and Adr1 transcription factors known to bind the WWCCCCRC cis element in yeast. Length of the boxes is proportional to number of amino acids. C Graphical summary illustrating how variation in cis-regulatory regions (gray boxes) leads to differential activation (colored plain arrowheads) of fungal genes (orange and green boxes) and could have contributed to the evolution of virulence on Arabidopsis in Sclerotinia. Hexagons and circles are fungal proteins produced by genes of the corresponding colors. The block arrow indicates functional inhibition. TF transcription factor.

Full size image

Discussion

We analyzed the global transcriptome of S. sclerotiorum during the colonization of hosts from six botanical families, providing a unique opportunity to test for the existence of generalist and host-specific transcriptomes in this fungal pathogen. While previous investigations revealed adaptations to a true generalist lifestyle supporting the colonization of any host [28, 29], we provide here molecular evidence for polyspecialism, the use of multiple independent modules dedicated to the colonization of specific hosts. These findings indicate that adaptation to new hosts can select for generalist and polyspecialist features within a single genome. We highlight a key role of regulatory variation in conserved fungal genes for the expansion of host range in this pathogen lineage.

We identified a subset of host species triggering specialized transcriptome reprogramming in S. sclerotiorum. Genes related to detoxification of host defense compounds were enriched in the specialized transcriptomes, while the core transcriptome overrepresented functions associated with carbohydrate catabolism and sugar transport. Host-specific regulation of pathogen genes confers the ability to quantitatively modulate the virulence program to match requirements for specific hosts. Transcriptional plasticity contributes to successful colonization of different host plants in aphids and a range of hemi-biotrophic fungal and oomycete pathogens [24, 61]. Host-specialized transcriptomes are often small and consist of secreted proteins with roles in modulating host plant defense responses and nutrient assimilation [61] as for the head blight pathogen F. graminearum [62, 63] and the septoria leaf blotch pathogen Zymoseptoria tritici [64]. A host-specialized transcriptome has been suggested previously for S. sclerotiorum when infecting B. napus and lupin (Lupinus angustifolius), although the majority of induced genes were induced on both hosts [65]. Our analysis on six dicot host species revealed that 52% of S. sclerotiorum genes upregulated in planta were host-specific. Predicting how the relative proportion of host-specific transcripts would differ by including more hosts in the analysis remains challenging since the number of host-specific induced genes varied considerably according to host (from 0% on S. lycopersicum to 36.9% on B. vulgaris). Clear host-specific gene regulation suggests that transcriptome-based reverse ecology approaches are feasible, allowing for instance to identify new host defense mechanisms based on pathogen transcriptome data.

Gene losses and gene gains by recombination, horizontal gene transfer, or copy number variation are efficient means to modify the host range [66, 67]. For instance, horizontal gene transfer contributed to host range expansion of the Metarhizium genus [22]. Also, gene copy number variation is often driven by repetitive and transposable elements and contributes to pathogenicity and can provide high variability of diverging paralogs of effectors, for example in powdery mildew fungi [68]. These mechanisms can enable rapid adaptation to new hosts and contribute to host range expansion in some lineages. Nevertheless, gene presence/absence polymorphism and coding sequence changes can have detrimental effects, in particular in unstable environments. For instance, pathogen gene loss may be advantageous on host carrying a matching R protein, but detrimental otherwise. Because coding but not cis-regulatory mutations are sensitive to frameshift, coding sequence variation is more likely to be detrimental and pleiotropic. In addition, transcription factor binding sites are short in comparison with coding sequence and therefore more likely to be neofunctionalized. This is the reasoning behind the cis-regulatory hypothesis stating that mutations that alter the regulation of gene expression are more likely to contribute to phenotypic evolution [69]. In agreement with this, we provide evidence that cis-regulatory variation contributes to the evolution of camalexin responsiveness in Sclerotinia. S. trifoliorum is closely related to S. sclerotiorum but has only been reported on plants from the Fabaceae family and rare cases on hosts from the Asteraceae and Plantaginaceae [55, 70]. All Sclerotiniaceae species and strains we tested were highly sensitive to camalexin, except for S. sclerotiorum strains. In spite of a high level of similarity in the genome of S. sclerotiorum strain 1980 and S. trifoliorum strain SwB9, their transcriptomes upon camalexin treatment appeared clearly distinct. Our transcriptomic analysis on camalexin focused on the respective highest tolerable concentration for each of the two fungi in order to ensure a comparable physiological state amid camalexin pressure. S. sclerotiorum generally displays little host preference [71] and further investigations will be required to determine the extent of variation in camalexin sensitivity and transcriptomic response to camalexin at the intraspecific level. Our findings highlight regulatory variation as an adaptive strategy for fungal pathogens jumping to new host plants. We propose that a certain degree of transcriptional plasticity or stochasticity for genes encoding promiscuous enzymes, conserved effectors, and their regulators, enable pathogen survival as endophytes on non-host plants. The persistence of a founder endophyte population may lead to the fixation of alleles with an adaptive expression pattern enabling the emergence of a pathogenic lifestyle on new hosts. Importantly, this implies that the gene pool the common ancestor of both Sclerotinia species was pre-equipped to adapt on non-host plants, and that a certain degree of fluctuations and randomness in gene regulation could enable the emergence of new traits. In this scenario, host range expansion does not require the acquisition of new genes from horizontal gene transfer or outcrossing. In the context of the impact of plant pathogens on food security, this observation calls for monitoring not only crop pathogen populations but also their close relatives infecting wild species in the same environment.

Our current analysis suggests that promiscuous enzymes with a flexible transcriptional pattern in Sclerotinia mediate the detoxification of host defense metabolites in the appropriate context. In addition, conserved “core” effectors with conserved plant targets could enable host jumps or host range expansions (e.g. [2]), followed by or in parallel with transcriptional plasticity. We discovered that the 70 S. sclerotiorum genes induced on camalexin and A. thaliana contain one predicted effector (Sscle_05g046060) and 23 secreted proteins (Dataset S1: Table S13). Among them, 22 genes are conserved in S. trifoliorum SwB9, 20 of which are conserved in over 100 fungal species, probably being part of a core secretome. Current functional information is not sufficient to determine whether these genes harbor a conserved function, contribute to Sclerotinia virulence, or act on plant antifungal chemicals. Polymorphisms in the amino acid sequence of core secreted proteins could lead to functional innovations enabling for instance the detoxification of a broader range of compounds in the context of host range expansion and speciation, providing a complementary adaptive mechanism to the regulatory variation uncovered in this work.

The current study focused on S. sclerotiorum infection of A. thaliana and relationship to its typical secondary metabolites, indole glucosinolates and camalexin, owing to the tractability of this system. We also found brassinin to be more toxic on S. trifoliorum than on S. sclerotiorum. S. sclerotiorum uses the brassinin glucosyltransferase SsBGT1 (Sscle01g003110) to detoxify this defense compound [59], a gene conserved in S. trifoliorum (SCLTRI_002931). In our assays, SsBGT1 was induced on all plants except S. lycopersicum. S. sclerotiorum SsBGT1 and its S. trifoliorum ortholog were induced to similar levels on camalexin (LFC 1.31 and 1.27 respectively). This data made brassinin a less straightforward determinant of Sclerotinia host range than camalexin. Nevertheless, future studies on the evolution of brassinin detoxification in the Sclerotiniacea should provide insights into host adaptation in these fungal pathogens.

Besides A. thaliana, S. sclerotiorum exhibited clear host-specific transcriptomes on castor bean, sugar beet, and common bean. On castor bean and sugar beet, genes related to the detoxification of host compounds were enriched in the specific transcriptomes. This may reflect the specific response to phytotoxins common to these host plants, such as the red beet antimicrobial phenolic secondary metabolites known as betalains [72] and the ribosome-inactivating protein ricin from castor bean [73]. By contrast, tomato, sunflower, and common bean induced limited host-specific transcriptomes. Flavonoids are common antimicrobial phytotoxins produced by many vascular plants and abundant in sunflower and beans for example [74]. While there are more than 10,000 flavonoid structures, S. sclerotiorum may use a conserved pathway to tackle these toxins. Indeed, the quercetin dioxygenase SsQDO (Sscle07g059700) catalyzes the cleavage of the flavonol carbon skeleton, thus targeting a range of flavonoids [75].

Our promoter region analyses identified a motif enriched in the cis-elements of S. sclerotiorum genes but not their orthologs in S. trifoliorum. In baker’s yeast, the motif is recognized by the zinc-finger transcriptional regulators ADR1 and Mig1-3, which have two likely orthologues in Sclerotinia and Botrytis. Mig1-3 are orthologous to the Aspergillus carbon catabolite repressor CreA, which regulates plant cell wall-degrading enzymes in A. nidulans [76] and human mycosis disease in A. fumigatus [77]. A. flavus creA mutants are defective in aflatoxin biosynthesis and crop colonization [78]. In the apple blue mold pathogen Penicillium expansum, CreA acts as positive regulator of the mycotoxins patulin and citrinin, and P. expansum creA mutants are near-avirulent on apples [79]. The S. sclerotiorum SsCreA (Sscle01g002690) could therefore be involved in the tight host-specific regulation of mycotoxin biosynthesis and plant cell wall-degrading enzymes. The targets of these transcription factors in S. sclerotiorum could direct the appropriate response to plant defense compounds, including activating the detoxification machinery for glucosinolates. In addition, DNA-binding proteins may have acquired the capability of binding the WWCCCCRC motif in S. sclerotiorum. For example, the putative C2H2 fungal transcription factor Sscle07g059580 was induced both by camalexin and on A. thaliana and might have evolved a specific role in the response to Brassicaceae. The evolution of camalexin responsiveness may therefore have involved cis- and trans-regulation. Further analysis of the transcription factors SsCreA, SsADR1, and Sscle07g059580 regarding their binding efficacy of the WWCCCCRC motif, the repertoire of their genomic targets, and their specific function in response to Brassicaceae is required to answer these questions.

Our findings reveal that host range expansion can be supported by regulatory variation in genes conserved in related non-adapted fungal pathogen species. The genetic requirements for such an evolutionary trajectory need to be determined in order to identify nonpathogenic species with a high potential for jumping to new hosts through regulatory variation. One explanation could be that core effectors, effector innovation, and regulatory features contribute to various degrees to host jumps and expansions. This mechanism enables the emergence of new disease with no or limited gene flow between strains and species, and could underlie the emergence of new epidemics originating from wild plants in agricultural settings.