Constitutive expression of transcription factor SirZ blocks pathogenicity in Leptosphaeria maculans independently of sirodesmin production

Sirodesmin, the major secondary metabolite produced by the plant pathogenic fungus Leptosphaeria maculans in vitro, has been linked to disease on Brassica species since the 1970s, and yet its role has remained ambiguous. Re-examination of gene expression data revealed that all previously described genes and two newly identified genes within the sir gene cluster in the genome are down-regulated during the crucial early establishment stages of blackleg disease on Brassica napus. To test if this is a strategy employed by the fungus to avoid damage to and then detection by the host plant during the L. maculans asymptomatic biotrophic phase, sirodesmin was produced constitutively by overexpressing the sirZ gene encoding the transcription factor that coordinates the regulation of the other genes in the sir cluster. The sirZ over-expression strains had a major reduction in pathogenicity. Mutation of the over-expression construct restored pathogenicity. However, mutation of two genes, sirP and sirG, required for specific steps in the sirodesmin biosynthesis pathway, in the sirZ over-expression background resulted in strains that were unable to synthesize sirodesmin, yet were still non-pathogenic. Elucidating the basis for this pathogenicity defect or finding ways to overexpress sirZ during disease may provide new strategies for the control of blackleg disease.


Introduction
Leptosphaeria maculans is the most significant biological threat to the global canola industry [1]. This fungal pathogen displays a complex hemibiotrophic life cycle, which begins with an initial biotrophic infection of cotyledons or young leaves followed by lesion formation. The fungus travels via the petiole into the stem, and down the plant to the base of the stem where it causes cankering, and hence the common name of the disease blackleg [2].

Growth of L. maculans strains, pathogenicity and competition testing
L. maculans strains were cultured in 10% cleared V8 juice (Campbell's) that had been adjusted to pH 6 using NaOH prior to autoclaving (CV8). 2% agar was added to produce solid media. Cultures on solid media were kept at 22˚C with a 12 h light-dark cycle to induce sporulation while liquid cultures were grown at 22˚C in the dark. The wild type strain used was D5. This strain was isolated in Penshurst, Victoria, in 1988 and is part of the International Blackleg of Crucifers Network collection (strain IBCN18, also called M1). The ability of strains to cause disease was conducted on canola Brassica napus cv. Westar or B. juncea cv. Aurea, with lesion areas measured 14 days post inoculation, as described previously [18]. These are two Brassica cultivars routinely used for assessing blackleg symptoms; the seed used were obtained from previous bulking production in the laboratory. Antimicrobial activity in the strains was tested against bacterium Bacillus subtilis and fungus Leptosphaeria biglobosas as described in detail previously [12].

RNA-sequencing expression analysis
Previously generated RNA-sequencing data were used to examine the expression of genes in the sirodesmin biosynthesis cluster. As in Urquhart and Idnurm [19], raw RNA-sequencing reads from two studies compareing L. maculans transcript levels in vitro and in planta [14,15] were obtained from the sequence read archive at NCBI, and mapped to the L. maculans strain JN3 genome sequence [20] using Geneious version 11 software. The TPM (Transcripts Per Kilobase Million) values were calculated for the 3´exon of each gene to minimize bias effects of long transcripts.

Generation of strains to over-express sirZ during plant disease
The recombinant strains created in this study, starting in the wild type strain D5, are listed in Table 1. Oligonucleotide primers used to make constructs for transformation were synthesized by Sigma-Aldrich, Australia, with the sequence details provided in S1 Table. A construct to constitutively express SirZ was developed by cloning the coding region of the sirZ gene, amplified using primers AU226 and AU227, into the BglII site of plasmid PLAU2 [21] to produce plasmid PLAU50. This resulted in the sirZ gene being expressed under the control of the L. maculans actin promoter. Similarly, genes encoding green fluorescent protein (GFP) and the avirulence protein AvrLm6 were amplified using primer pairs AU24 and AU27; and AU28 and AU31, respectively, and cloned into the BglII site of PLAU2 for the transformation of L. maculans to create control strains. Cloning used Gibson assembly and subsequent transformation into Escherichia coli strain NEB 1 5-alpha (New England Biolabs, USA). The constructs were transformed by electroporation into Agrobacterium tumefaciens strain EHA105 as described previously [18]. Strain EHA105 is a commonly used host for subsequent transformation of plants and fungi [22]. Fungal transformation was conducted on Agrobacterium induction medium, followed by an overlay with 200 μg/ml of hygromycin and 200 μg/μl of cefotaxime in CV8 medium [23]. Multiple independent transformants were isolated onto selection medium, and then assessed for their pathogenicity on canola.

Generation of gene disruption strains by CRISPR-Cas9 mutations
One of the SirZ over-expression strains (6B) was selected for the construction of additional mutations. To enable genes in the sirodesmin biosynthesis pathway to be disrupted in this strain using the CRISPR-Cas9 system it was first transformed with an Agrobacterium-mediated T-DNA from plasmid pMAI23 that expresses a modified form of the Streptomyces pyogenes Cas9 gene [21], using selection on 100 μg/ml G418. Three guide RNA constructs, for targeting the Cas9 endonuclease to the L. maculans sirP, sirG and sirZ genes, were produced. These were made by cloning the oligonucleotides "SirP", "SirG" and "SirZ" (sequences in S1 Table) into the XhoI site of plasmid pMAI75 by Gibson assembly (NEBuilder HiFi Assembly Cloning Kit, New England Biolabs), and then by subcloning the gRNA constructs using restriction enzymes NheI and SpeI into the XbaI site of the plasmid pPZPnat1 [24], which confers resistance to nourseothricin when the T-DNA is transformed into L. maculans.
The T-DNAs that express the CRISPR guide RNA constructs were transformed into a sir-ZOE-Cas9 strain, using 100 μg/ml nourseothricin as the selective agent. In this system, incorrect repair of the Cas9 endonuclease damage introduces mutations into the target regions. Transformants were screened by PCR and restriction enzyme digests to identify those with mutations at the target sites of the three genes. Transformants were then cultured through a step of isolating a colony from a single spore to ensure nuclear homogeneity, and the gene regions amplified and analyzed using Sanger sequencing of PCR amplicons to define the mutations.
Extraction of secondary metabolites, and their separation by thin layer chromatography or high performance liquid chromatography 25 ml CV8 liquid were inoculated with approximately 10 6 spores of L. maculans strains. After culturing for seven days at 22˚C in darkness, mycelia were removed by filtration through miracloth and the culture filtrate solution (volume approximately 20 ml) was extracted once with 20 ml ethyl acetate. The extracts were dried under a stream of nitrogen gas and the residue suspended in 200 μl pure ethanol.
For thin layer chromatography (TLC), extracts were resolved on silica gel 60 F 254 aluminium sheets (Merck) using a 1:1 ratio of chloroform to ethyl acetate, and visualized using UV light, as in [25].
For high performance liquid chromatography (HPLC), 10 μl of each sample was injected into an Agilent 1100 machine that was run with an H 2 O-acetonitrile gradient using a Synergi 4u Fusion-RP 80A column (250 × 4.60 mm; C18; Phenomenex) as described previously [4]. Molecules eluting from the column were detected by absorbance of 240 nm UV light.
Mass spectrometry was used to confirm the identity of relevant peaks on the HPLC chromatogram. Samples peaks were manually collected and analyzed on a Thermo Q Exactive Orbitrap mass spectrometer in the positive mode.

The genes for sirodesmin biosynthesis are down-regulated during the foliar stages of blackleg disease
If a protein is important for a function, its gene should be expressed when needed: two previously published RNA-seq data sets [14,15] were re-examined for the expression of the sir gene cluster. All 18 annotated genes in the cluster are down-regulated during pathogenesis on cotyledons relative to in vitro at all time points. For example, the in planta data of Sonah et al. at 11 dpi or Lowe et al. at 14 dpi on cotyledons, conducted in two different countries, show the same pattern of low gene expression of the sir genes (Fig 2).
Mapping the transcriptomic data back to the genome sequence of L. maculans revealed two additional genes within the sir cluster that were not previously identified, which we name sirK and sirX. The predicted protein sequence for SirK shows homology to GliK encoded in the Aspergillus fumigatus gene cluster and gliK is essential for the synthesis of gliotoxin [26]. SirX does not show clear homology to any proteins in GenBank based on BLAST searches. Both sirK and sirX are also down-regulated during the early stages of disease, consistent with the trend of the other sir genes. The sequences of the two genes have been deposited in GenBank as accessions MK609857 and MK609858.

The L. maculans actin promoter can be used to drive protein expression in planta
If the sir genes are down regulated during infection, the question was what would happen if they were expressed constitutively in strains? A candidate promoter region is actin, used in numerous gene expression studies as a standard for constitutive and strong gene expression.
To assess the ability of the L. maculans actin promoter to drive expression of proteins during early pathogenesis on cotyledons, two proteins were expressed; green fluorescent protein (GFP) and the avirulence factor AvrLm6. GFP was clearly visible as green fluorescence in vitro (not shown) and in planta (Fig 3A). Expression of AvrLm6 resulted in a loss of pathogenicity on Brassica juncea cv. Aurea, which contains the Rlm6 resistance gene that recognizes AvrLm6 (Fig 3B), while the wild type, which has a mutation in the AvrLm6 avirulence gene, causes lesions. The avirulence genes are amongst those most highly expressed during the early stages of infection relative to in vitro, and hence these results indicate that using the actin promoter has the ability to drive expression to the levels able to impact plant disease.

Expression of transcription factor SirZ during the early biotrophic disease stage results in a loss of pathogenicity
All five independent transformants expressing the actin promoter-sirZ construct that were tested showed a loss of pathogenicity phenotype (Fig 4). To rule out a possible inadvertent gene silencing effect of the construct, secondary metabolites were extracted from culture filtrates of the untransformed wild type and one over-expression strain (D5+sirZOE#6), resolved by thin layer chromatograph, and visualized under ultraviolet light. The over-expression strain still produced sirodesmin (data not shown). Next, to establish if the loss of pathogenicity is due to the synthesis of sirodesmin or some other effect, three genes required for sirodesmin synthesis were targeted for mutation in one representative of the over-expression strain background (i.e. strain D5+sirZOE#6). CRISPR-Cas9 constructs were transformed into this strain, and candidate mutations identified by PCR analysis. Sanger sequencing of the gene regions in these sirZ, sirP and sirG mutants made in the SirZ overexpressing background identified frameshift mutations indicative of loss-of-function alleles (Fig 5).

PLOS ONE
Over expression of SirZ prevents blackleg disease Disruption of the introduced sirZ allele via CRISPR-Cas9 mutation resulted in a restoration of pathogenicity (Fig 4). On the other hand, disruption of either the sirP or sirG genes, that encode enzymes for the synthesis of sirodesmin (Fig 1), did not restore pathogenicity.
Sirodesmin has other activities, such as an inhibitor of bacterial and fungal growth. We therefore tested the strains for their ability to inhibit growth of bacterium Bacillus subtilis or close relative of L. maculans, Leptosphaeria biglobosa. All strains tested maintained some level of inhibition of bacterial or fungal growth, and which was variable between different experiments (data not shown).

Over-expression of SirZ results in a decreased growth rate that does not account for the loss in pathogenicity
The five actin promoter-sirZ transformants showed a slight reduction in growth rate and darkening of colony color (Fig 6). Disruption of sirZ as well as sirG and sirP reversed this effect  https://doi.org/10.1371/journal.pone.0252333.g006 (Fig 6), suggesting that this phenotype is attributed to the production of sirodesmin. Because the growth defect is reverted in the sirG or sirP mutants, yet these strains remain non-pathogenic, this indicates that the loss of pathogenicity caused by sirZ overexpression is not simply due to slower growth.

HPLC analysis of secondary metabolites in wild type and mutant strains confirms gene functions
HPLC analysis revealed that the sirZ and sirP mutants produced neither sirodesmin nor its precursor phomamide, thus confirming the function of these genes had been abolished ( Fig  7A). The sirG mutant produced phomamide but not sirodesmin, and the sirG mutant produced a number of additional products. This is consistent with the position of where the enzymes SirP and SirG act in the pathway for sirodesmin synthesis (Fig 1). Mass-spectra obtained from selected HPLC fractions confirmed that the 11.5 min peak represents phomamide and the 17.5 min peak represents sirodesmin PL (Fig 7B).

Discussion
The transcription factor SirZ, which regulates the synthesis of the major secondary metabolite produced by L. maculans [8,9], was constitutively expressed and this blocked pathogenicity in cotyledon infection assays (Fig 4). Because the strain is no longer pathogenic, the transcript levels of sirZ itself during infections could not be measured due to the limited amount of fungal biomass. However, as a control the same promoter used to drive expression of GFP or the avirulence factor AvrLm6 both led to the functional expression of high levels of these proteins. Disruption of the SirZ overexpression allele, via CRISPR/Cas9, restored pathogenicity, indicating that the loss of pathogenicity was related to the expression of SirZ.
In vitro the SirZ constitutive expression strains had slower radial growth rates compared to the wild type, which may reflect autotoxicity (Fig 6). However, this growth rate is not the cause of the reduction in pathogenicity. First, L. maculans wild isolates exhibit a range of growth rates yet cause equal lesion sizes on cotyledons (A.P. Van de Wouw, pers. commun). Second, we have isolated mutants with vegetative growth defects yet remain fully pathogenic [19]. Third, disruption of either sirP or sirG resulted in restoration back to wild type growth rate but not to wild type pathogenicity. Hence, the overexpression of sirZ leads to a true pathogenicity defect rather than a general reduction in strain fitness.
To explore the mechanism behind the loss of pathogenicity when sirZ is overexpressed, we selected two genes in the sirodesmin biosynthesis cluster for disruption using CRISPR/Cas9: sirP and sirG. The sirP gene was chosen because it has previously been shown to be required for sirodesmin production and is expected to act early in the biosynthetic pathway, which agrees with the fact we did not detect sirodesmin or phomamide production in the sirP mutant. The sirG gene was chosen because this gene has not previously been demonstrated to be essential for sirodesmin production, although homology to gliG in the gliotoxin biosynthetic pathway of A. fumigatus suggested that it was responsible for adding the sulfur atoms to the molecule by acting downstream of phomamide [6]. The phytotoxicity of sirodesmin (as well as other epipolythiodioxopiperazines) has been attributed to the disulfide bridge that enables these toxins to cross-link proteins via cysteine residues [27]. The HPLC chromatograph of the sirG disrupted strain supports such a role as phomamide, but not sirodesmin, was present in the culture filtrates. In addition to phomamide, a number of novel peaks were observed in the sirG mutant. This is similar to the gliG deletion strain of A. fumigatus that produced compounds not observed in either the wild type or gliP (homologous to L. maculans sirP) deletion strain [6].
Given that the strains in which sirG and sirP were mutated in the sirZ overexpression background remained non-pathogenic we conclude that the loss of pathogenicity due to SirZ overexpression is not due to the over-production of sirodesmin. Two possible explanations for this unexpected finding are that expression of one of the sirodesmin biosynthesis pathway genes is causing the loss of pathogenicity or that SirZ regulates additional targets outside of the cluster. The possibility of the second explanation is supported by the fact that regulatory cross talk between biosynthetic gene clusters has been reported previously in other fungi. A number of variations have been observed. These can be situations in which a transcription factor can regulate two gene clusters, as for ScpR and the RsmA-AflR regulon in Aspergillus nidulans [28,29] or FapR in A. fumigatus [30]. Some transcription factors within a cluster can have widespread influence on secondary metabolism, such as Tri6 in Trichoderma arundinaceum that regulates numerous secondary metabolite clusters although this may also be via alteration of the mevalonate biosynthetic genes [30]. The consequences of overexpression can thus be complex, e.g. if a transcriptional regulator also influences the availability of the substrates for secondary metabolites. Or, another curious situation is the transcription factor HasA that regulates hexadehydroastechrome synthesis yet also controls genes involved in iron homeostasis. The consequences of overexpression of HasA, most likely mediated via changes in metal homeostasis, then impacts the expression of 14 gene clusters in addition to the has cluster [31]. One final example worth noting is the case of the synthesis of azophilones, which requires the activities of two separate gene clusters in Aspergillus terreus. In this case, a transcription factor in one of the clusters acts to induce the expression of two other transcription factors one each found in the two clusters [32]. In light of these examples, in either the case of another sir gene or secondary metabolite being involved, these findings from L. maculans are another caution against assuming that phenotypes due to overexpression of putative pathway specific transcription factors can be solely attributed to increased metabolite production.
Discovering which genes, other than those in the sirodesmin gene cluster, are regulated when sirZ is over-expressed should reveal new genes required for disease development. A bioinformatic analysis of the genes in the epipolythiodioxopiperazine synthesis gene clusters from three species revealed a consensus motif of TCGGNNNCCGA is in the promoters of eight of the L. maculans sir genes [8], and it is hypothesized that SirZ binds to this motif. We therefore examined the secondary metabolite core enzymes and surrounding associated genes in the L. maculans genome, as defined in the MycoCosm database [33], by searching manually for the relaxed motif of CGGNNNCCG. The genome is predicted to encode 13 non-ribosomal peptide synthases, 12 polyketide synthases, and two terpene synthases [20]. Eight motifs were found in the sir cluster, as previously reported [8]. A single candidate motif was associated with six other gene clusters. Four weaker candidates were in the gene promoter of the putative major facilitator superfamily transporter for the 'NPS3 cluster', between a fatty acid transporter and hypothetical protein in the NPS11 cluster, the monooxygenase in the NPS2 cluster, and between the major facilitator superfamily transporter and NPS1. The two most conserved motifs were for the monooxygenase in the PKS1 cluster and between a hypothetical protein and NPS4. None of these gene clusters have defined functions. Hence, the identification of other SirZ-regulated genes may require RNA-sequencing comparisons or could also be addressed by a mutant screen seeking a restoration to wild type pathogenicity.
Sirodesmin is the best characterized secondary metabolite of L. maculans, and yet this study further complicates its possible functions in the biology of this and other fungi. The metabolite is produced during in vitro growth and in stem lesions, yet sir gene expression is turned down during the early stages of plant infection. These findings tend to support a hypothesis that the fungus produces the metabolite for competition with other microbes as part of its saprotrophic growth. Sirodesmin has other activities, such as inhibition of bacteria or fungi [12,34]. The strains created here all still maintained the ability to inhibit bacterial or fungal growth. However, these types of studies on living cultures, rather than using purified metabolites, can be problematic in cases whereby bacterial-fungal or fungal-fungal interactions trigger the production of secondary metabolites potentially in both organisms [35]. Similarly to L. maculans, the 'primary' functions of other ETPs have been difficult to assess. For instance, gliotoxin was isolated as an antifungal agent, is immunosuppressive, has anti-tumor properties and in specific conditions contributes to the virulence of A. fumigatus (the inconsistent findings are reviewed in [36]). Roles for gliotoxin are well illustrated in Trichoderma virens, where the gliP mutant loses its mycoparasitic activity against the oomycete Pythium ultimum and the fungus Sclerotinia sclerotiorum, and is less pathogenic against the insect Galleria mellonella [37].
This study suggests a promising new direction for future studies on L. maculans on the investigation of genes that are not normally expressed during disease. Most previous work on L. maculans has focused on either avirulence genes or on pathogenicity genes, which are two groups of genes that are normally expressed during disease. However, RNA-sequencing data [14,15] show that a large number of genes are expressed at a high level during in vitro growth but at a very low level during the early biotrophic stages of disease. Despite having been probed for a role in pathogenicity [4,8,12,24] the genes of the sirodesmin biosynthetic cluster fall into this group. It is possible that many such genes are not only dispensable at this stage of disease but moreover that their expression during this stage would be deleterious. We have demonstrated this to be the case for sirZ. If strategies can be developed to cause pathogenic fungi to miss-express such genes this would be a powerful tool in the fight against plant disease.
Supporting information S1 Table. Oligonucleotide primers used in this study. (PDF)