The host generalist phytopathogenic fungus Sclerotinia sclerotiorum differentially expresses multiple metabolic enzymes on two different plant hosts

Sclerotinia sclerotiorum is a necrotrophic fungal pathogen that infects upwards of 400 plant species, including several economically important crops. The molecular processes that underpin broad host range necrotrophy are not fully understood. This study used RNA sequencing to assess whether S. sclerotiorum genes are differentially expressed in response to infection of the two different host crops canola (Brassica napus) and lupin (Lupinus angustifolius). A total of 10,864 of the 11,130 genes in the S. sclerotiorum genome were expressed. Of these, 628 were upregulated in planta relative to in vitro on at least one host, suggesting involvement in the broader infection process. Among these genes were predicted carbohydrate-active enzymes (CAZYmes) and secondary metabolites. A considerably smaller group of 53 genes were differentially expressed between the two plant hosts. Of these host-specific genes, only six were either CAZymes, secondary metabolites or putative effectors. The remaining genes represented a diverse range of functional categories, including several associated with the metabolism and efflux of xenobiotic compounds, such as cytochrome P450s, metal-beta-lactamases, tannases and major facilitator superfamily transporters. These results suggest that S. sclerotiorum may regulate the expression of detoxification-related genes in response to phytotoxins produced by the different host species. To date, this is the first comparative whole transcriptome analysis of S. sclerotiorum during infection of different hosts.


Results
Adapter trimming and BBSplit read assignment. RNA sequencing was used to compare the gene expression of S. sclerotiorum during infection of the host species L. angustifolius and B. napus, along with an in vitro control treatment. RNA was extracted from infected plant samples at 3 days post-inoculation (dpi) and sequenced with the Illumina HiSeq platform. The Trimmomatic package was used to remove adapters resulting from the sequencing platform, along with low-quality reads 25 . Trimmomatic read retention appeared to be generally higher in the in vitro treatments than in the in planta treatments, with respective averages of 95.0% and 92.4% (Table 1). BBSplit was used on the resulting trimmed reads to assign reads to either S. sclerotiorum or the respective plant host (B. napus or L. angustifolius). The proportion of reads mapped to S. sclerotiorum varied considerably between the three treatments, from approximately 7.5% (in B. napus) to 83.9% (in L. angustifolius) (Fig. 1). The mean proportion of reads assigned to the S. sclerotiorum genome across all libraries was 39.3%.
Data inspection, quality control and sample clustering analysis. The Limma package in R was used to test differential gene expression across the entire S. sclerotiorum genome, making all three possible comparisons between treatments 26,27 . As expected, S. sclerotiorum-derived reads were detected in all treatments. Between the three treatments, a total of 10,864 predicted protein-encoding genes were detected at least once, representing approximately 97.6% of the protein-encoding genome. Average gene coverage was greatest in the in vitro treatment at approximately 96.9% and slightly lower in the in planta treatments at 92.1% in B. napus and 94.9% in L. angustifolius.
Principal component analysis plots and heatmaps produced in Limma demonstrated a consistent grouping of the in vitro samples (Fig. 2). However, the L. angustifolius and B. napus samples did not form distinct groups. The Limma PCA plot indicated that one of the B. napus replicates was distinct from the others based on its gene expression profile ( Fig. 2A). Subsequently, a heatmap produced in Limma indicated that there were in fact two outlying replicates, with one from each in planta treatment (Fig. 2B) and statistical adjustments were made for the outliers as described in the materials and methods. In both cases, the in vitro samples were relatively consistently clustered.
Differential expression analysis reveals a set of Sclerotinia sclerotiorum genes highly up regulated on two different host species. We hypothesised that the majority of in planta-upregulated genes would be induced irrespective of host. To test this hypothesis, we compared the sets of genes upregulated by S. sclerotiorum in B. napus and L. angustifolius relative to the in vitro treatment. We did not count genes with less than 1 read per million (RPM) in all three replicates and considered genes differentially expressed if they had a  Table 1. Quality control and assignment rates from Trimmomatic and BBSplit. The fifth column refers to the number of in planta S. sclerotiorum reads as a proportion of all in planta reads assigned to the host or pathogen by BBsplit. log 2 (fold change) of more than 2 and an adjusted p value of <0.05. In total, 628 genes were significantly upregulated in at least one of the in planta treatments relative to in vitro (Fig. 3A,B, Supplementary Table 1). The majority of these genes (64%) were upregulated in both in planta treatments, supporting our hypothesis. In both in planta treatments, the upregulated gene with the largest log-fold change relative to in vitro was sscle_15g106470 (14.6 in B. napus (p adj = 0.004) and 14.8 in L. angustifolius (p adj = 0.003)), which encodes a thioesterase (Table 2). In terms of absolute expression levels, the most highly expressed upregulated gene was sscle_16g108170 in both species, with log-counts-per-million (LCPM) values of 13.7 in B. napus (p adj < 0.001) and 13.4 in L. angustifolius (p adj < 0.001). This gene encodes for a predicted glycoside hydrolase. Necrotrophic phytopathogens may utilise a combination of secreted effectors, secondary metabolites and CAZymes to facilitate infection 3,20,28,29 . Therefore, we hypothesised that genes within these functional categories would be over-represented among genes upregulated in planta relative to in vitro. To test this hypothesis, a Chi-square test of independence (α = 0.05) was conducted to assess the association between upregulation and gene function, for the categories of CAZyme-associated genes, secondary metabolite-associated genes and putative effector-encoding genes. Using DBCan2, we identified 173 CAZymes in the S. sclerotiorum genome. Using AntiSMASH, we identified 87 secondary metabolite biosynthesis clusters containing 1,630 genes. To identify effector genes, we used the 70 effector predictions from Derbyshire et al. (2017).
Gene function and upregulation were found to be associated (X 2 = 311.99, df = 3, p < 0.001). All three mentioned gene categories were upregulated more frequently than expected by chance alone. Notably, 98 CAZyme-encoding genes and 20 secondary-metabolite associated genes were upregulated, which is considerably greater than the expected frequencies of approximately 27.7 and 9.5 respectively. expression analysis shows that Sclerotinia sclerotiorum differentially regulates a subset of 53 genes on two different host species. Expression analysis conducted in Limma showed that S. sclerotiorum differentially regulated a subset of 53 genes between the two host species. These included 21 genes upregulated in L. angustifolius relative to B. napus, with the remaining 32 downregulated (up-regulated on B. napus relative to L. angustifolius) (Fig. 3C). Again, we did not count genes with less than 1 read per million (RPM) in all three replicates and considered genes differentially expressed if they had a log 2 (fold change) of more than 2 and an adjusted p value of <0.05.
The in planta differentially expressed genes included secreted CAZymes and a solitary putative effector (Fig. 4). The five differentially expressed CAZymes included three glycoside hydrolases, a pectin lyase fold and a www.nature.com/scientificreports www.nature.com/scientificreports/ vanillyl-alcohol oxidase (Fig. 4A). All but one of these CAZymes were upregulated in L. angustifolius relative to B. napus (Fig. 4B). The differentially expressed putative effector had no associated GO terms. None of the putative secondary metabolite genes were differentially expressed on the two different hosts (Fig. 4C).
Quantitative polymerase chain reaction supports the hypothesis differential expression of a subset of genes between host environments. To further test the hypothesis that some genes are specifically up-regulated in S. sclerotiorum in response to host environment, we performed quantitative PCR (qPCR). The 15 genes we subjected to qPCR are listed in Supplementary Table 2, along with their expression profiles in the RNA sequencing analysis. These genes included nine that, according to the RNA sequencing analysis, were significantly up-regulated in B. napus relative to L. angustifolius and six vice versa, and two that were up-regulated in both in planta samples relative to in vitro.
Our first experiment (experiment 1) was a set of duplicate (L. angustifolius) and triplicate (B. napus and in vitro) samples independent from the RNA sequencing samples but generated in the same fashion. Out of the eight genes analysed in this experiment, seven showed expression changes in the same direction as they did in the RNA sequencing analysis (Fig. 5). The two genes sscle_05g040340 and sscle_08g06410 in particular showed convincing evidence of up-regulation on B. napus relative to L. angustifolius as they both exhibited non-overlapping standard errors between the two samples. According to the RNA sequencing data, the genes sscle_15g106480 and sscle_15g106510 were up-regulated in both L. angustifolius and B. napus relative to in vitro. Our qPCR assay agreed with this, however, the standard error between the two replicates was high for the L. angustifolius samples and it overlapped with that of the in vitro samples. The gene sscle_02g012440 did not exhibit the same expression profile as was observed in the RNA sequencing data. In the RNA sequencing analysis, this gene was up-regulated in B. napus relative to L. angustifolius. However, in our qPCR analysis, we observed the opposite.
To formally test the hypothesis of differential expression between conditions, we performed Welch's t-test for B. napus and L. angustifolius comparisons, and analysis of variance when considering all three samples, B. napus, L. angustifolius and in vitro. For a single gene, sscle_05g040340, we could confidently accept the alternate hypothesis of differential expression between two samples based on qPCR data (B. napus: mean 2 −∆Ct = 3.46, standard deviation +/−3.51; L. angustifolius: mean 2 −∆Ct = 0.045, standard deviation +/−0.035. Welch's t = 4.3, df = 2.89, P = 0.025). We infer that the general agreement between qPCR and RNA sequencing data suggests that the RNA sequencing results would likely be similar across further replicated data sets. However, we can only confidently state that a single candidate gene has strong evidence for up-regulation on B. napus relative to L. angustifolius.
In our second experiment (experiment 2), we infected detached leaves using mycelial matts. This experiment was replicated four times per condition. We tested nine genes in this experiment, two of which were also included in experiment 1 (the genes sscle_01g006290 and sscle_05g040340). Overall, seven out of nine genes exhibited changes in expression between B. napus and L. angustifolius in the same direction as observed in the RNA sequencing data; these included the two genes shared between experiment 1 and experiment 2. However, in contrast to experiment 1, in experiment 2, these two genes exhibited overlapping standard errors of their mean 2 −∆Ct values. Four out of the seven genes had non-overlapping standard errors, suggesting strong differential expression. These included sscle_08g062510, sscle_05g047210 and sscle_01g005000, which were up-regulated on B. napus and sscle_04g033880, which was up-regulated on L. angustifolius. These all agreed with the direction of change in expression in the RNA sequencing analysis. The two genes that disagreed with the RNA sequencing data were sscle_02g018820 and sscle_05g040320. The former was expressed to a greater level on B. napus in the qPCR data and L. angustifolius in the RNA sequencing data. The latter showed only a negligible increase in expression from L. angustifolius to B. napus, whereas in the RNA sequencing analysis this difference was substantial.
To formally test the hypothesis of differential expression between conditions in experiment 2, we performed Welch's t-tests. For two genes, sscle_08g062510 and sscle_05g047210, we could relatively confidently accept the alternate hypothesis of differential expression between the two samples based on qPCR data (sscle_08g062510: B. napus: mean 2 −∆Ct = 1.15, standard deviation = 0.9; L. angustifolius: mean 2 −∆Ct = 0.38, standard deviation = 0.15. Welch's t = 2.59, df = 4.71, P = 0.052. sscle_05g047210: B. napus: mean 2 −∆Ct = 0.7, standard deviation = 0.3; L. angustifolius: mean 2 −∆Ct = 0.19; standard deviation = 0.06. Welch's t = 3.71, df = 4.66, P = 0.016). We infer that the general agreement between qPCR and RNA sequencing data suggests that the RNA sequencing results would likely be similar across further replicated data sets. However, we can only confidently state that two candidate genes have strong evidence for up-regulation on B. napus relative to L. angustifolius.

Discussion
Necrotrophic phytopathogens such as S. sclerotiorum are known to utilise a variety of pathogenicity factors to facilitate infection, including CAZymes, proteinaceous effectors and secondary metabolites 7,20,30 . Existing evidence on pathogenicity factors in S. sclerotiorum consists of a combination of qPCR and RNA-seq based expression studies, knockout assays and biochemical studies focussing on select genes 13,20,23 . These studies include investigations of the broad-spectrum effector SsSSVP1, the metabolite oxalic acid, and a group of 16 putative effectors 20,21,23 . To our knowledge, this study is the first to utilise RNA sequencing to investigate host-specific gene expression across the entire S. sclerotiorum transcriptome, including a wide range of predicted pathogenicity factors. Many of the upregulated genes do not appear to have been examined by previous host-specific expression studies, with the exception of a selection of putative effectors 23 .
According to the results of a Chi-square test, secreted CAZymes, secondary metabolites and putative effectors were collectively upregulated in planta relative to in vitro (X 2 = 311.99, df = 3, p < 0.001). These results suggest that these groups of genes are indeed involved in the infection strategy of S. sclerotiorum. Secondary metabolites, effectors and CAZymes have previously been described in more than one necrotrophic species. Notable examples include the secondary metabolite botrydial in B. cinerea 7 , the effector SsSSVP1 in S. sclerotiorum 20 and a variety of predicted CAZymes including cellulases, pectinases and xylanases in S. sclerotiorum 31 . Consequently, the upregulation of these gene categories by S. sclerotiorum is consistent with earlier work implicating them in the infection process.
Among the main groups of pathogenicity factors, CAZymes formed the largest group of upregulated genes. A total of 98 secreted CAZymes were upregulated by S. sclerotiorum in planta relative to in vitro in at least one species. A variety of CAZymes have been predicted in the S. sclerotiorum genome, which include pectinases, glucanases and cellulases 4 . Many of these genes are thought to encode plant cell wall-degrading enzymes (PWDEs), which are involved in degrading the diverse structural molecules of plant cell walls such as cellulose,  Table 3. All genes significantly differentially expressed by S. sclerotiorum between the in planta treatments. "Abs. LFC" refers to the magnitude of log-fold change in gene expression between the in planta treatments (base 2), "LCPM" refers to the gene expression level in log-counts-per-million (base 2). "L" refers to the L. angustifolius treatment, "C" refers to the B. napus treatment.  4 . The upregulation of secreted CAZymes in planta is therefore entirely consistent with their theoretical role in the necrotrophic infection strategy.
Putative secondary metabolites were also upregulated by S. sclerotiorum in planta relative to the in vitro treatment. Though some secondary metabolites are thought to act as secreted toxins in other fungal necrotrophs [32][33][34][35] , they appear to be poorly understood in S. sclerotiorum. By comparison, the secondary metabolites botrydial and botcinic acid are known to be important virulence factors in B. cinerea 7 . A variety of polyketide synthases (PKS) and non-ribosomal peptide synthases (NRPS) are encoded within the S. sclerotiorum genome, suggesting that the fungus may produce polyketides and non-ribosomal peptides as secondary metabolites 13 . These compounds are known to be among the phytotoxic secondary metabolites secreted by the closely related B. cinerea 28 .
Two PKS-encoding genes, namely sscle_15g106480 and sscle_15g106510, were upregulated in planta in this study in at least one plant treatment. The orthologues of these genes in B. cinerea are involved in the production of botcinic acid, which is an important pathogenicity factor in the species 7 . Limited evidence is available regarding the production of botcinic acid by S. sclerotiorum, and the compound has not been characterised in this pathogen 7 . Seifbarghi et al. 13 found that only sscle_15g106480 was upregulated by S. sclerotiorum during infection of B. napus, and suggested that the pathogen is unlikely to produce botcinic acid because both sscle_15g106480 and sscle_15g106510 are required for synthesis 7 . In this study, both genes were upregulated in L. angustifolius, which is consistent with the synthesis of botcinic acid or a related compound (LFC = 3.8, p adj < 0.001, LFC = 9.2, p adj < 0.001 respectively). However, in B. napus these genes were significantly downregulated relative to in vitro, which suggests that the role of botcinic acid may be host specific (LFC = −3.3, p adj < 0.001, LFC = −5.8, p adj < 0.001 respectively). Perhaps the downregulation of these genes in B. napus suggests that the plant defence responses have been largely overcome, negating the need to secrete botcinic acid. However, it is difficult to draw any convincing conclusions given the advanced stage of sample infection and the absence of evidence on the production of botcinic acid by S. sclerotiorum 7   www.nature.com/scientificreports www.nature.com/scientificreports/ The primary aim of this study was not to investigate gene expression in planta relative to in vitro, but to examine differential gene expression as a response to different plant hosts. In comparison to the 628 genes upregulated in planta, only 53 genes were differentially expressed between the two plant hosts (B. napus and L. angustifolius) (Fig. 3c). The relatively small number of host-specific differentially expressed genes (DEGs) is consistent with the PCA and hierarchical clustering analysis, which consistently identified differences between the in vitro and in planta treatments but failed to identify distinct B. napus and L. angustifolius treatments. One explanation for the small number of significantly differentially expressed genes is that the environments presented by the plant hosts to S. sclerotiorum are more similar to one another than they are to the in vitro control. Alternatively, perhaps the number of differentially expressed genes does not accurately reflect the differences between the plant hosts, and both species present biological cues that prompt upregulation of broad-spectrum virulence factors in planta. Some effectors could potentially be upregulated in all plant hosts if they interact with highly conserved plant genes. The S. sclerotiorum-derived effector SsSSVP1 is a potential example of this effect as it is thought to act on mitochondrial respiration 20 . Accordingly, this gene was not significantly differentially expressed between the plant hosts in this study.
It is also worth considering the possible influence of the late sampling stage on the results. The infection process of S. sclerotiorum is thought to be a two-step process, in which an early biotrophic phase is followed by a later necrotrophic phase 13,14 . S. sclerotiorum is thought to secrete oxalic acid and possibly other compounds as "compatibility" factors to subvert host defence responses during the early biotrophic stage of infection 13,31 . Differential expression of genes associated with these compatibility factors has been suggested as a form of host-specific adaptation in S. sclerotiorum 13 . Any compatibility factors differentially expressed during the earlier biotrophic stage of infection would not necessarily be represented in this study, as the biotrophic phase of infection is thought to occur 12-24 hours post-inoculation (hpi), which is considerably earlier than the 72 hpi time point used in this study 13 .
Seifbarghi et al. 13 suggested two products in particular as potentially host-specific compatibility factors, neither of which were shown to be differentially expressed in this study. The first of these is secreted integrin-like  13 . The y axis shows the inverse ∆Ct value relative to the housekeeping gene Sclerotinia sclerotiorum β-tubulin. The x axis shows the names of genes included in the analysis. The gene names coloured in green correspond with the RNA sequencing analysis, whereas those in red do not. The thick horizontal black lines represent median values. The boxes and whiskers represent interquartile range and the black points represent outliers. For most of the genes, we were concerned with differential expression between Lupinus angustifolius and Brassica napus, as this was the primary hypothesis we were testing. We also selected the two genes sscle_15g106480 and sscle_15g106510 as they were significantly upregulated on both B. napus and L. angustifolius relative to in vitro. (2019) 9:19966 | https://doi.org/10.1038/s41598-019-56396-w www.nature.com/scientificreports www.nature.com/scientificreports/ protein SSITL, which suppresses jasmonic acid/ethylene-mediated resistance responses 13,36 . The associated gene sscle_08g068500 was not significantly differentially expressed by S. sclerotiorum between the two plant hosts. The second of these candidate host-specific compatibility factors is the chorismate mutase SScm1 13 . This enzyme has not been characterised, but is thought to be secreted by S. sclerotiorum to suppress salicylic acid synthesis, which is a key plant defence response 37 . The gene responsible for SScm1 has not been positively identified, though the gene sscle_16g111080 matches the domain 38 . This gene was not significantly differentially expressed by S. sclerotiorum between L. angustifolius and B. napus, suggesting that it is not involved in the transcriptional adaptation of S. sclerotiorum to these host species in the late stages of infection. However, we emphasise here that the late time points of our study preclude a comprehensive overview of gene expression at all stages of infection in both hosts. The discrepancies between the work of Seifbarghi et al. and us may be better explained by methodological differences. Additionally, in our experiment, one of the samples generated from infected B. napus tissue produced relatively few reads mapping to the S. sclerotiorum genome. This low abundance of mapped reads led to significant divergence of the replicate from others in the dataset. This discrepancy could have also affected inference from our differential expression analysis.
The production of salicylic acid may be vulnerable to interference at more than one stage. Earlier research in Ustilago maydis suggests that SA production could be reduced by fungal secretions of chorismate mutase, which could potentially be produced by the gene sscle_16g111080 38,39 (Table 3). Another potential target is isochorismate, which follows chorismate in one of the SA synthesis pathways in plants 40 . The gene sscle_15g106860, which has an isochorismatase-like Pfam domain, was significantly upregulated in L. angustifolius relative to B. napus (LFC = 4.6, p adj = 0.026). Isochorismatases catalyse the breakdown of the isochorismate into smaller by-products, potentially diverting isochorismate from subsequent stages of the synthesis process. This may suggest that sscle_15g106860 is involved in the disruption of SA synthesis in a host-specific fashion.
A total of five predicted CAZyme-associated genes were differentially expressed by S. sclerotiorum between the two host species. Of these five genes, all but one encoded for glycoside hydrolases. Glycoside hydrolases are a highly diverse group of enzymes that catalyse the hydrolysis of glycosidic bonds between carbohydrate subunits, reflecting the structural variety of their substrates 41 . The remaining gene, sscle_02g021040, encoded a predicted vanillyl alcohol oxidase and was upregulated in L. angustifolius relative to B. napus (LFC = 2.2, p adj = 0.038). It is difficult to speculate on the role of this gene, as the role of vanillyl alcohol oxidases in fungi is generally poorly understood 42 .
The differentially expressed glycoside hydrolases (GHs) may hint at differences between the composition of plant cell walls. The four predicted GHs all belonged to different GH families, without any obvious pattern. The families GH12, GH3, GH28 and GH16 are associated with the degradation of cellulose, beta-glycans, pectins 30 and hemicellulose respectively 43 (Table 3). These substrates are all well-known cell wall components across the plant kingdom, which raises the question of why certain CAZymes would be upregulated in specific hosts 44 . One possibility is that these differentially expressed CAZymes act on variants of cell-wall components peculiar to specific plant hosts. The cell-wall components hemicellulose and pectin are known to vary considerably in chemical composition and structure between plant species, which provides some support for this proposition 45 . This diversity is further reflected in the S. sclerotiorum genome, which contained some 215 GH-associated genes according to HMMER predictions conducted in DbCan2. However, it is difficult to be specific about the roles of these host-specific CAZyme-associated genes without functional characterisation. The few CAZyme genes characterised by Li et al. 6 do not correspond to the gene models developed by Amselem et al. 4 or Derbyshire 38 , and consequently it is unclear whether these CAZymes correspond to differentially expressed genes in this study. Despite the inability to assign specific roles to CAZyme-associated DEGs, the results of this study indicate a role for gene expression in the adaptation of S. sclerotiorum to different host substrates.
Host-specific expression of effector candidate genes was more restricted than may be expected based on previous research. Of the effector candidates predicted by Derbyshire et al. 38 and Amselem et al. 4 , only one effector candidate gene, namely sscle_08g064180, was significantly differentially expressed between the two host species. This gene was upregulated in B. napus relative to L angustifolius (LFC = 2.4, p adj = 0.007)( Table 3). By contrast, a previous quantitative PCR study of a select group of 16 effector candidates in S. sclerotiorum produced five separate groups of effector candidate genes, each with varying patterns of expression across time points and species 23 . The differentially expressed putative effector sscle_08g064180 is predicted to have a coiled coil domain, though there appears to be little evidence regarding the role of these domains in effectors.
Though secreted CAZymes, effectors and secondary metabolites play a heavily emphasised role as pathogenicity factors in the necrotrophic infection strategy, they may not tell the full story of host-specific gene expression. Previous research into host-specific gene expression in fungal necrotrophs has focussed on pathogenicity factors such as effector candidates and CAZymes to the exclusion of others 22,23 . As a result, the broader story of host-specific gene expression is poorly understood. In this study, the 53 genes differentially expressed by S. sclerotiorum in a host-specific fashion included only six encoding for CAZymes, putative effectors and secondary metabolites, which are often considered important factors in molecular plant-necrotroph interactions 7,20,22 . Investigation of the remaining genes is necessary to provide a broader picture of host-specific S. sclerotiorum-host interactions.
A total of 47 host-specific DEGs were not associated with CAZymes, putative effectors or secondary metabolites. What then, is the function of this surprisingly large group of genes? One possibility is that some of the host-specific DEGs are involved in the detoxification of phytotoxic compounds secreted by the plant host. S. sclerotiorum, for example, is known to metabolise phytoalexins produced by crucifer hosts during infection, effectively converting them into smaller non-toxic molecules 8 . Similarly, the fungal phytopathogen Grosmannia clavigera, which infects pines, is thought to utilise a variety of mechanisms to detoxify host-produced terpenes 9 . These mechanisms may include modification of the terpenes to less toxic molecules and transporter-based efflux 46 . Flexible detoxification mechanisms in S. sclerotiorum could contribute to its remarkable adaptability to diverse host species. www.nature.com/scientificreports www.nature.com/scientificreports/ Several host-specific DEGs are possible candidates for involvement in the detoxification of plant-produced secondary metabolites. Of particular interest are the three host-specific cytochrome P450-encoding DEGs, sscle_01g006290 (LFC = 4.1, p adj < 0.001), sscle_04g033880 (LFC = 4.2, p adj = 0.002) and sscle_08g067130 (LFC = 3.3, p adj = 0.002)( Table 3). Cytochrome P450s (CYPs) are a broad group of genes that are commonly involved in metabolism pathways, and are particularly well known for the modification and degradation of xenobiotic compounds 47,48 . Examples of CYPs in fungi include the CYP53 family, which are known to be involved in detoxifying isoeugenol, benzoic acid, and other phytotoxic compounds produced by plant hosts in Cochliobolus lunatus 49 . The differential expression of CYPs by S. sclerotiorum in a host-specific manner, with log-fold changes as great as 4.2, suggests that these genes may be expressed in response to species-specific plant-derived compounds. Given the known role of CYPs in the detoxification of plant secondary metabolites by phytopathogens, it is quite possible that some of the CYP-associated DGEs may be involved in detoxification.
A variety of other host-specific DGEs could be involved in fungal detoxification of plant molecules. The genes sscle_08g067140 (LFC = 2.0, p adj = 0.036) and sscle_10g076570 (LFC = 2.5, p adj = 0.004) appear to encode for tannases, which degrade tannins (Table 3). Plant-produced tannins are known to be toxic to fungal phytopathogens such as Crinipellis perniciosa and Pythium insidiosum 50,51 . Consequently, it appears likely that the differential expression of tannase-encoding genes indicates detoxification of plant-produced tannins. Interestingly, sscle_08g067140 and sscle_10g076570 were upregulated in different host species (L. angustifolius and B. napus respectively), suggesting that the plant hosts produce different forms of tannin.
The metallo-beta-lactamase encoding gene sscle_05g040340 may also play a role in the detoxification of plant secondary metabolites by S. sclerotiorum (LFC = 6.8, p adj < 0.001)( Table 3). In other species, metallo-beta-lactamases are well-known as metabolic enzymes of xenobiotic compounds. Notably, the New Delhi metallo-beta-lactamase-1 enzyme produced by several human-pathogenic bacteria is known to be involved in the metabolism of antibiotics, resulting in antibiotic resistance 52 . In fungi, these metallo-beta-lactamases are thought to act on a range of lactams not restricted to beta-lactams 10 . Lactones, which are chemically related to lactams, are known to be produced by plants 53 . Several synthetic lactones have been shown to have antifungal effects on fungal pathogens including B. cinerea, Penicillium citrinum, and several Aspergillus species 53 . Consequently, it has been suggested that fungal metallo-beta-lactamases may be involved in the degradation of toxic secondary metabolites produced by plant hosts 10 . In this case, this may suggest that S. sclerotiorum is upregulating a metallo-beta-lactam-associated gene for the purpose of detoxifying a lactone produced by B. napus. Though these S. sclerotiorum-plant detoxification interactions have not been characterised, the several differentially expressed domains associated with secondary metabolite modification hint at the possibility of secondary metabolite degradation as a means of interspecific adaptation in S. sclerotiorum.
Thee major facilitator superfamily-domain genes were differentially expressed by S. sclerotiorum in a host-specific fashion, namely sscle_02g012440 (LFC = 3.9, p adj = 0.004), sscle_05g040320 (LFC = 5.3, p adj = 0.001) and sscle_05g047210 (LFC = 3.2, p adj = 0.004) ( Table 3). MFS proteins are involved in the active transport of compounds across cell membranes, and are present in both eukaryotes and prokaryotes 54 . Some MFS proteins are known to export toxic xenobiotic compounds and have been implicated in the resistance of organisms to antibiotics, fungicides and phytotoxins 55,56 . One possibility is that the differentially expressed MFS genes are involved in the detoxification of host-specific phytotoxins. Relatively few membrane transporters are known to be involved in the detoxification of phytotoxins such as phytoalexins, and most of these transporters are ATP-binding cassette ABC transporters 11 . One exception is the MFS-encoding gene MgAtr5, which is thought to be involved in the efflux of the phytotoxins resorcinol and resveratrol in Mycosphaerella graminicola during infection of wheat 12 . The detoxification role played by MgAtr5 suggests that MFS-based efflux of phytotoxins could be possible in S. sclerotiorum.
This study investigated host-specific gene expression in the fungal necrotroph S. sclerotiorum between the host species B. napus and L. angustifolius. A total of 628 genes were upregulated in planta, including a number of secreted CAZymes, putative effectors and secondary metabolite-encoding genes. Of arguably greater interest were the smaller set of 53 genes that were differentially expressed between the two host species. Many of these genes had potential roles in the detoxification of plant-derived secondary metabolites.
Fungal detoxification of phytotoxins has been previously investigated in relation to chemical control of fungal pathogens. New fungicide chemistries could aid the plant's natural defence responses by inhibiting fungal detoxification-related enzymes, preventing the metabolism of phytotoxins. Further research into detoxification pathways in fungal pathogens could provide additional targets for similar fungicides, potentially improving the variety of fungicides available for controlling pathogens such as S. sclerotiorum.

Materials and Methods
Biological materials. B. napus cv. 'Cobbler' and L. angustifolius cv. 'Tanjil' plants were grown from seed in a plant growth chamber with a 16 hour photoperiod and 16 °C/22 °C temperature cycle. Relative humidity was maintained at 60% and the daytime light intensity was 200 µmol/m 2 /s. S. sclerotiorum cultures (isolate CU11. 19) were prepared from dry sclerotia on potato-dextrose agar (PDA) 57 . After two days of growth, mycelial plugs were taken from the actively growing edges of the PDA cultures and subcultured onto minimal glucose medium.
For RNA sequencing and qPCR experiment 1, plants were inoculated at 5 weeks post-sowing with 5 mm minimal media agar plugs taken from the actively growing edges of the S. sclerotiorum cultures. The plugs were bound to the stem with Parafilm to conserve moisture. The in vitro control consisted of S. sclerotiorum cultures cultivated in 50 mL of potato-dextrose broth. These in vitro cultures were inoculated with 5 mm minimal media agar plugs. For qPCR experiment 2, detached leaves of B. napus and L. angustifolius plants were inoculated with mycelial matts grown in minimal medium as per Seifbarghi et al. (2017) 13 .
The plant sections, mycelial matts and in vitro cultures were harvested at 3 days post-inoculation (DPI). Stem sections were cut from the plants at the outer extent of visible tissue necrosis. The plant samples were immediately Differential expression analysis and quality assessment. Differential expression analysis was conducted in order to determine which genes were significantly differentially expressed between the two hosts, and the in vitro sample. The Limma (v3.38.1) package in R was used for this analysis because it has the capacity to compare more than two treatments in a single stage of analysis 26,27 . During the initial screening of the gene expression results, genes with counts-per-million (cpm) values of less than 1 in more than 3 samples were removed, because they were unlikely to be significantly differentially expressed. The remaining read counts were normalised using the "TMM" method.
The differential expression profiles of the samples were visualised using the plotMDS() function from Limma and the heatmap.2() function from gplots 65 . According to these plots, one B. napus replicate (C2) and possibly one L. angustifolius replicate (L3) appeared to be have a very distinct expression profile from the other replicates. To correct for the influence of these outliers, the voomWithQualityWeights() function was used to weight replicates based on their similarity. Pairwise comparisons were made between all three experimental treatments (L. angustifolius in planta, B. napus in planta and in vitro). All comparisons were made at the α = 0.05 significance level after false discovery rate (FDR) correction, and only log-fold changes (LFCs) greater than 2 were considered biologically significant.
Pathogenicity factors and DbCan2 secreted CAZyme prediction. CAZymes, secondary metabolites and proteinaceous effectors are known pathogenicity factors in necrotrophic pathogens such as S. sclerotiorum. DbCan2 (v7) was used to predict putative CAZymes from the S. sclerotiorum genome 66 . The S. sclerotiorum strain 1980 protein sequences were used as the input for CAZyme prediction (GCA_001857865) 38 . The putative CAZYmes predicted by DbCan2 were filtered before use in further analysis. Only CAZymes predicted by at least two of the three algorithms employed by DbCan2 were retained, in order to limit the number of spurious CAZyme predictions. To focus on secreted CAZymes, the predicted CAZymes were restricted to those with positive SignalP results. Putative effectors were obtained from Derbyshire et al. and Amselem et al. 4,38 . Secondary metabolite genes were identified using AntiSMASH.
Quantitative polymerase chain reaction. The 15 genes in Supplementary Table 2 were analysed using quantitative PCR (qPCR) as a complementary approach to RNA sequencing for detection of differential expression between samples. We performed two experiments, experiment 1 consisted of samples generated in exactly the same way as for the RNA sequencing experiment. For this experiment, there were three replicates for in vitro and B. napus derived samples and two for L. angustifolius derived samples. All samples were independent of those used for RNA sequencing, so they represent further replication of this analysis. Experiment 2 was performed using methods adapted from Seifbarghi et al. (2017) 13 . In this experiment, mycelial matts of S. sclerotiorum were first grown on minimal medium inoculated with a PDA plug. The matts were transferred to detached leaves of B. napus or L. angustifolius and removed for RNA extraction at 3 DPI. Experiment 2 was replicated four times for each condition.
The RNA extracted from these samples was converted to cDNA using the Roche first strand cDNA synthesis kit for RT-PCR (AMV). The cDNA samples were then diluted 1/20 before qPCR. The qPCR analysis was performed using the Bio-Rad iTaq Universal SYBR Green Supermix according to the manufacturer's instructions. The primers used are detailed in Supplementary Table 3. The thermocycler settings were 95 °C for 2 min, then 95 °C for 15 sec, 60 °C for 30 sec and 72 °C for 15 sec, repeated 40 times, followed by 72 °C for 2 min. Data were