Assignment of a dubious gene cluster to melanin biosynthesis in the tomato fungal pathogen Cladosporium fulvum

Pigments and phytotoxins are crucial for the survival and spread of plant pathogenic fungi. The genome of the tomato biotrophic fungal pathogen Cladosporium fulvum contains a predicted gene cluster (CfPKS1, CfPRF1, CfRDT1 and CfTSF1) that is syntenic with the characterized elsinochrome toxin gene cluster in the citrus pathogen Elsinoë fawcettii. However, a previous phylogenetic analysis suggested that CfPks1 might instead be involved in pigment production. Here, we report the characterization of the CfPKS1 gene cluster to resolve this ambiguity. Activation of the regulator CfTSF1 specifically induced the expression of CfPKS1 and CfRDT1, but not of CfPRF1. These co-regulated genes that define the CfPKS1 gene cluster are orthologous to genes involved in 1,3-dihydroxynaphthalene (DHN) melanin biosynthesis in other fungi. Heterologous expression of CfPKS1 in Aspergillus oryzae yielded 1,3,6,8-tetrahydroxynaphthalene, a typical precursor of DHN melanin. Δcfpks1 deletion mutants showed similar altered pigmentation to wild type treated with DHN melanin inhibitors. These mutants remained virulent on tomato, showing this gene cluster is not involved in pathogenicity. Altogether, our results showed that the CfPKS1 gene cluster is involved in the production of DHN melanin and suggests that elsinochrome production in E. fawcettii likely involves another gene cluster.


Introduction
Secondary metabolites (SMs) are compounds produced by microbes, plants and insects that are often repurposed as medicines and pesticides. Equally important are SMs with harmful effects, such as mycotoxins and pathogenicity factors that poison animals or promote crop diseases. The vast majority of fungal SMs with a clear biological role in situ are pathogenicity or PLOS  laccases [34]. Although the DHN melanin genes can be present in fungal genomes, the DOPA melanin pathway is the major route employed by certain fungal species, as exemplified by the pine needle pathogen Dothistroma septosporum [35]. Cladosporium fulvum is a non-obligate, biotrophic fungus that causes tomato leaf mold disease. C. fulvum shows limited filamentous growth on in vitro media in the dark, forming small sporulating colonies. They exhibit a green-brown colour, which was linked to the production of the pigment cladofulvin [36]. C. fulvum colonies harbour a grey colour when cladofulvin is not produced, which is likely due to the production of another pigment [36]. C. fulvum is known to reproduce asexually only and production of cladofulvin is primarily observed in conidia [37]. Despite a high potential chemical diversity with 23 predicted-functional SM core genes [38,39], the pigment cladofulvin produced by the claG gene cluster remains the only detectable SM [36,40]. It was suggested and later shown that the repression of cladofulvin biosynthetic genes is required for biotrophic growth of C. fulvum [37,38]. CfPKS1 is another nrPKS core gene that shows a similar expression profile during infection of tomato leaves, i.e. downregulation [38]. CfPKS1 belongs to a predicted gene cluster containing genes that encode a prefoldin chaperone (CfPRF1), a reductase (CfRDT1) and a transcription factor (CfTSF1) [38]. Previous comparative genomic analyses indicated that the CfPKS1 gene cluster is homologous to the Elsinoë fawcettii gene cluster responsible for elsinochrome production, a lightactivated toxin involved in the virulence of this pathogen on citrus hosts [41,42]. However, the phylogeny of CfPks1 suggested that it is also closely related to nrPKSs involved in DHN melanin biosynthesis [38]. In another study, CfPKS1 was strongly up-regulated in the C. fulvum deletion mutant Δcfwor1 during growth on agar [43]. The hyper-black appearance of Δcfwor1 colonies and the absence of detectable SMs suggested that CfPks1 might be involved in the production of polymerized DHN melanin in C. fulvum. Such ambiguity between elsinochrome and DHN production remains unresolved.
Here, we report the functional characterization of the CfPKS1 gene cluster by targeted gene deletion (CfPKS1), over-expression of the predicted local regulator (CfTSF1), and heterologous expression in Aspergillus oryzae. We provide chemical evidence of the pigment produced by this pathway and assessed the role of this compound in pathogenicity and biotrophic growth of C. fulvum.

Definition of the CfPKS1 gene cluster in Cladosporium fulvum
The CfPKS1 gene cluster ( Fig 1A) was initially predicted solely through its homology and synteny with the characterized elsinochrome gene cluster in E. fawcettii, as only minimal gene expression within this gene cluster had been observed during the growth of wild-type C. fulvum under diverse conditions [38]. The gene cluster includes CfTSF1, a gene predicted to encode a pathway-specific transcription factor [38,41]. To up-regulate and clearly define the CfPKS1 gene cluster, wild-type C. fulvum was transformed with a plasmid containing CfTSF1 fused to the promoter region of the nitrogen-regulated C. fulvum Avr9 gene [44]. The resulting OE.CfTSF1 transformant (S1 Fig) does not show any in vitro difference compared to wild type, but this transformant is not pathogenic on tomato (S2 Fig). Although random insertion of the expression cassette in a pathogenicity gene cannot be excluded, this loss of pathogenicity is likely due to the up-regulation of CfTSF1 because the Avr9 promoter induces high-expression when C. fulvum enters the plant and colonizes leaf tissues [38,44]. Both transformant and parental strain were grown in PDB and then induced in B5 medium without nitrogen (B5-N) for 48 hours to induce gene expression. Transcriptional profiling by RT-qrtPCR showed that the relative expression of CfPKS1, CfTSF1 and CfRDT1 was 1.7, 14.6 and 46.5-fold higher, respectively, in the OE.CfTSF1 transformant than in wild type (each t-test P-value < 0.0001), whilst the predicted border genes are not co-regulated ( Fig 1B). In contrast, CfPRF1 was not co-regulated and is therefore unlikely part of the CfPKS1 biosynthetic pathway. The up-regulation of the CfPKS1 gene cluster is specific to the over-expression of CfTSF1 because the gene  [38,41]. The locus of the DHN melanin gene cluster in Alternaria alternata and Cochliobolus heterostrophus is shown for comparison [32,33]. All loci contain a prefoldin-encoding gene downstream of the polyketide synthase gene. CMR1 and CMRA encode transcription factors that regulate the DHN melanin biosynthetic genes. BRN1 and BRM2 are 3-hydroxy naphthalene reductases homologous to RDT1. Black arrows indicate co-regulated genes in each species, white arrows indicate non-co-regulated genes and the grey triangle indicate a transposable element. Loci are not drawn to scale. (B) Relative expression of genes at the CfPKS1 locus determined by reverse transcription-quantitative real-time polymerase chain reaction (RT-qrtPCR). Strain were pre-cultured in Potato Dextrose Broth (PDB) for five days before the biomass was transferred to B5 without nitrogen (B5-N) medium. After 48 hours, the biomass was recovered and used for RNA isolation and cDNA synthesis. The expression value for each gene within the CfPKS1 locus was measured in wild-type C. fulvum and inducible over-expression tranformants C. fulvum OE. CfTSF1 and OE.CfClaE [38] grown in B5-N medium. Expression values were normalised to actin and the average value was plotted with standard deviation between three biological replicates. A two-way ANOVA with a posthoc Sidak multicomparison test at the significance level of 0.05 was used to calculate statistically significant changes in gene expression between wild type and transformant strains. Asterisks ( � ) denote statistically significant changes (p < 0.05 or less). (C) Differential expression of genes at the CfPKS1 locus and homologues of genes involved in the biosynthesis of DHN melanin in the Δcfwor1 deletion mutant compared with the wild type as determined by RNA-seq [43]. Black and gray bars show significant and nonsignificant fold changes, respectively, according to Cuffdiff analysis of three biological repeats.
In a previous study, the CfPKS1 core gene was found to be highly up regulated in Δcfwor1 deletion mutants [43]. The published RNA-seq data of this mutant confirms that all genes from the predicted gene cluster are co-regulated, but CfPRF1 and the predicted border genes are not (Fig 1C). Consistent with previous phylogenetic analyses of 4THN synthases [35,38,45], CfPKS1, CfRDT1 and CfTSF1 are all orthologous to genes involved in DHN melanin biosynthesis (Figs 2A-2C). In other fungi, this pathway involves three other genes, AYG1, 4HNR and SCD1 [12,34]. Orthologues of these genes were identified on different scaffolds in the genome of C. fulvum (Table 1; Figs 2C-2E) and all are significantly up-regulated in the Δcfwor1 deletion mutants (Fig 1C). In contrast, paralogues of these genes are not differentially expressed or are significantly down-regulated in the Δcfwor1 deletion mutants (Fig 1C).

CfPks1 is a polyketide synthase that releases 4THN
CfPKS1 is orthologous to other characterized 4THN synthases from fungi of different orders (Fig 2A). In C. lagenarium, the 4THN synthase ClPks1 carries a bi-functional TE domain that releases 4THN [18]. CfPks1 was expressed heterologously in A. oryzae M-2-3 in order to determine whether or not it catalyses the same reactions as ClPks1. Ethyl acetate extracts of transformants contained three major products, 1-3, bearing UV signatures diagnostic of aromatic polyketides (Fig 3). LC-MS analyses of organic extracts obtained from OE.CfTSF1 transformant after 48h growth on induction medium did not detect any precursor of DHN melanin. This observation could be due to a delay between the transcriptional induction and production of the compounds in significant amount as there was no pigmentation difference with wild type. Alternatively, it could suggest that DHN is efficiently polymerized, which is not easily extractable from cell walls with regular chemical methods.

CfPKS1 is needed for proper pigmentation of C. fulvum, but it is not required for pathogenicity on tomato
To confirm the heterologous expression results and obtain further insights in the biological role of the CfPKS1 gene cluster, the CfPKS1 gene was replaced by a deletion cassette containing the hygromycin resistance marker gene via homologous recombination (S1 Fig). Two confirmed independent deletion mutants and an ectopic transformant were selected for further analysis. Both Δcfpks1 deletion mutants were yellow-orange compared to the grey-green ectopic transformant and wild type ( Fig 4A). The wild-type strain shows similar coloration to the Δcfpks1 deletion mutants when it is grown in the presence of pyroquilon, an inhibitor of the DHN melanin pathway (Fig 4A; [7]). In contrast, it was not coloured differently when grown in the presence of hydroquinone, an inhibitor of DOPA melanin (Fig 4A; [7]). These results contrast with the situation in D. septosporum, a close relative species of C. fulvum that produces DOPA melanin despite the presence of DsPKS1, the orthologue of CfPKS1 and other 4THN synthase genes [35]. Together, these results show that the polymerization of DHN melanin contributes to the pigmentation of C. fulvum. The Δcfpks1 mutants do not manifest any other obvious developmental or physiological defects.
Because DHN melanin is a pathogenicity factor in several plant pathogens, we addressed the possibility that this role was also true in C. fulvum. Tomato plants were inoculated with wild-type C. fulvum, an ectopic transformant and two independent Δcfpks1 deletion mutants.

Fig 2. Phylogenetic analysis of proteins involved in DHN melanin biosynthesis.
Maximum likelihood phylogenetic trees were built to resolve the evolutionary relationship of (A) CfPks1 to characterized non-reducing polyketide synthases; (B) CfTsf1 to characterized regulators of DHN melanin biosynthesis; (C) CfRdt1 and CfRdt2 to reductases involved in the reduction of 1,3,8-trihydroxy-and 1,3,6,8-tetrahydroxy-naphthalene, respectively; (D) CfAyg1 to characterized hydrolases involved in polyketide deacetylation; and (E) CfScd1 to characterized scytalone dehydratases. Elsinoe fawcettii proteins assigned to the elsinochrome gene cluster are included [41]. Non-characterized homologues are also included for species in which the polyketide synthase involved in DHN synthesis is characterized. The name of characterized proteins and accession numbers (GenBank, SwissProt or Joint Genome Institute (JGI) protein ids) are indicated next to the species acronym. Clafu: Whilst the growth of all strains was similar at 4 and 8 dpi, the Δcfpks1 deletion mutants had significantly outgrown the control strains by 12 dpi (Fig 4B). By 16 dpi, this difference in growth became strikingly clear, as the white-orange deletion mutants had colonized a greater leaf surface area than the control strains ( Fig 4C).

The CfPKS1 gene cluster is involved in DHN melanin biosynthesis
A previous phylogenetic analysis suggested that CfPKS1 is an orthologue of EfPKS1, which encodes the nrPKS required for elsinochrome production [35,38,42]. The predicted elsinochrome gene cluster is also present in C. fulvum, except for the putative transporter ECT1  https://doi.org/10.1371/journal.pone.0209600.g003 [38,41]. However, the same phylogenetic analysis showed that CfPKS1 and EfPK1 are orthologous to 4THN synthases involved in the biosynthesis of DHN melanin. Similar observations were reported in other phylogenetic analyses [35,45]. Given that PKSs from a monophyletic clade tend to produce the same products, the different chemical structures of DHN melanin and elsinochrome contradict the phylogeny of their respective nrPKSs. The orthologous 4THN synthase in C. lagenarium has been characterized in detail; ClPks1 synthesizes the hexaketide 2-acetyl-1,3,6,8-tetrahydroxynaphthalene (ATHN) that is cyclized and deacylated by its bi-functional TE domain to give 4THN [18]. In this report, we showed that C. fulvum CfPks1 is also a synthase that releases 4THN because it produced both the hexaketide benzopyran and the pentaketide 4THN (Fig 3). The precursor ATHN was not detected in our experiments, which might be due to the presence of an expressed homologue of the hydrolase Ayg1 in A. oryzae that would efficiently deacetylate ATHN to yield 4THN [18,20].
In characterized nrPKSs involved in DHN melanin biosynthesis, aromatisation of the nascent ACP-bound polyketide to give a monocyclic intermediate is catalysed by the product template (PT) domain [14,16,18,46]. Closure of the second ring and release of the polyketide requires the Claisen/Dieckmann cyclase class of TE domains (TE/CLC) [14,18,46]. The absence of a functional TE domain results in the release of pyrone shunt products via spontaneous O-C cyclization [18]. The co-detection of 4THN and benzopyran in A. oryzae::CfPKS1 suggests that the TE domain of CfPKS1 was not fully functional, which might be due to the absence of tailoring enzymes. Ayg1 and its orthologues were shown to be involved in the deacetylation of a hexaketide or heptaketide precursor in the DHN melanin pathway described in blue/green fungi [24]. However, it was recently shown in the plant pathogen V. dahliae that AYG1 might also be important for DHN biosynthesis in brown/black fungi [47]. Our finding that CfAYG1 is present in C. fulvum genome and co-regulated with CfPKS1 suggests that this gene also plays a role in the DHN biosynthetic pathway in C. fulvum, which requires further investigation. Although another paralogue is present in C. fulvum, it is not co-regulated with CfPKS1 and thus is unlikely to be involved in 4THN biosynthesis (Figs 1 and 2). It is noteworthy that Yg1 in E. dermatitidis is actually a paralogue of characterized Ayg1 in other fungal species (Fig 2). The true orthologue of Ayg1 in E. dermatitidis must also be further investigated to ascertain its involvement in 4THN production.
Our phylogenetic analysis showed that co-regulated genes at the CfPKS1 locus in C. fulvum and genes from the predicted elsinochrome gene cluster (PKS1, RDT1 and TSF1) in E. fawcettii are orthologous to genes involved in DHN biosynthesis (Fig 2). A gene cluster for elsinochrome biosynthesis was recently characterized in Stagonospora nodorum [45]. This gene cluster was shown to be related to gene clusters involved in the biosynthesis of cercosporin, a compound of the perylenequinone family that is structurally related to elsinochrome [48][49][50]. Together with our characterization of CfPks1 as a 4THN synthase, these observations suggest that the elsinochrome gene cluster in E. fawcettii has not been accurately assigned. Instead, a cercosporin-like gene cluster is certainly involved in the production of elsinochrome as reported in S. nodorum. The fungus Cladosporium phlei produces phleichrome, a perylenequinone that is structurally related to elsinochrome and cercosporin [48]. The C. phlei nrPKS gene CpPks1, orthologous to CfPKS1 and EfPKS1, was assigned to phleichrome production [51]. Similarly, we suggest that CpPks1 is involved in DHN melanin production and another cercosporin-like nrPKS is responsible for phleichrome production. Such false assignments suggest that crosstalk and interdependencies between gene clusters might be more important than previously thought [52].

DHN melanin is not a virulence factor in C. fulvum
The biosynthesis of fungal DHN melanin has been extensively studied because of its diverse roles in fungal biology [15]. Melanin contributes to virulence in animal and plant pathogens, with the latter linked to the formation of host-invading appressoria [15]. In species that do not produce appressoria, melanin confers resilience to chemical and abiotic stresses [12]. C. fulvum does not differentiate appressoria, which likely explains why DHN melanin is not involved in the pathogenicity of this fungus. The pigmentation of C. fulvum relies on two compounds, the greyish DHN melanin and yellow-orange-purple (depending on pH) cladofulvin [36,40]. Cladofulvin is also not produced during plant infection, but instead protects the fungus from environmental stresses such as UV light and cold temperatures [37]. Although untested, DHN melanin likely plays a similar role in the survival of C. fulvum ex planta.
Although further investigation is required to fully exclude that the expression cassette has inserted within a pathogenicity gene, the observed loss of pathogenicity of the OE.CfTSF1 transformant suggests that adequate regulation and downregulation of DHN melanin production during leaf colonization is required for full pathogenicity in C. fulvum. In A. fumigatus and A. nidulans, melanin biosynthesis is initiated in endosomes that carry the enzymes that produce DHN, which is then polymerized within the cell wall by multicopper oxidase and laccase enzymes [53]. Melanin is polymerised in the cell wall, forming several layers of globular particles that grow thicker over time, strengthening the cell wall [54]. It is accepted that hyphal tip elongation requires enzymes to weaken the cell wall in order to incorporate new components, which are then cross-linked to rigidify the cell wall [55]. Thus, the abnormal accumulation of melanin in fungal cell walls is likely to modify its physical properties and increases its rigidity. We observed that the runner hyphae of C. fulvum OE.TSF1 transformants branched infrequently and the few successful penetration events resulted in colonizing hyphae that rapidly stopped growing (S2 Fig). This phenotype can be explained by the elevated production of DHN melanin and its abnormal accumulation in cell walls, which might reduce sensitivity to chemotactic gradients and then arrest growth of colonizing hyphae by preventing essential fungal tip remodelling [55]. An in planta analysis of fungal cell walls would address this hypothesis. Alternatively, DHN melanin accumulating in the cell wall could be recognized by plant cells, leading to the activation of plant defences and resistance. Similar in planta activation might be detrimental to the virulence of other plant pathogens irrespective of whether or not DHN melanin is a pathogenicity factor.

Conclusion
Using complementary approaches (gene expression, phylogeny, heterologous expression), our study confidently assigned the CfPKS1 gene cluster to DHN melanin in C. fulvum. It suggests that orthologous gene clusters in other species have been wrongly assigned to toxin production, including elsinochrome in E. fawcettii and phleichrome in C. phlei. Further investigations in these fungal species are needed to address this ambiguity and are likely to provide important insights on pathway crosstalk that might lead to incorrect gene cluster assignment.

Experimental procedures
Most of the methods were performed as described in Griffiths et al. (2016) and Griffiths et al. (2018) [37,40].

Fungal strains employed in this study
C. fulvum 0WU [39] was the parental strain used to perform transformation and gene deletion experiments. C. fulvum was grown on potato dextrose agar (PDA) plates at 20˚C in the dark. For inhibitor experiments, PDA medium was supplemented with 30 mg.L -1 pyroquilon or hydroquinone (Sigma-Aldrich, Zwijndrecht, The Netherlands). A. oryzae M-2-3 strain was used to perform heterologous expression [56].

Phylogeny
The protein sequence of CfPks1 and of selected characterized nrPKSs from groups II, III and IV [57] were aligned using Muscle [58] and poorly aligned regions of the alignment were removed using Gblocks, with half allowed gap positions for Ayg1, Rdt and Scd1 alignments, and allowing all gapped positions for the Tsf1 alignment [59]. Maximum-likelihood phylogeny was calculated using PhyML 3.1 [60] with the LG+G+I substitution model as determined by Modelgenerator v851 [61] and SH approximate likelihood ratio test to evaluate branch support.
Non-characterized homologues were retrieved from the Joint Genome Institute MycoCosm portal (genome.jgi.doe.gov; [62]) using BlastP (with default parameters) [63]. All Homologues of 4Hnr (AAG29497.2) and Sdh1 (BAA34046.1) from M. oryzae and of Ayg1 (AAF03354.1) from A. fumigatus were sought in the predicted proteome of C. fulvum. Homologues of CfRdt1, CfTsf1, CfRdt2, CfScd1 and CfAyg1 were sought in selected genomes. For each protein, retrieved homologues and characterized proteins were aligned using Muscle; sequences with large deletions or insertions were manually removed. Neighbour-Joining phylogenetic trees were built with the JTT substitution model using MEGA 7 [64] in order to ascertain orthology. Identified orthologues were then aligned again and analysed following the same process as for CfPks1 described above, but using the LG+G substitution model for Ayg1 and Scd1 trees.

Generation of OE.CfTSF1 and OE.CfTSF1::GFP transformants
The putative local regulator from the CfPKS1 gene cluster, CfTSF1, was amplified by PCR using Phusion Flash High-Fidelity PCR Master Mix (Life Technologies, Carlsbad, CA) from C. fulvum genomic DNA using the primer pair PacI_CfTSF1_Forward and NotI_CfTSF1_Reverse (S1 Table). Plasmid pFBTS3 contains the promoter of the nitrogen-regulated Avr9 gene [36,44]. The CfTSF1 amplicon and pFBTS3 were cut using PacI and NotI restriction enzymes (Fermentas Fast Digest, Waltham, MA), cleaned with Zymogen DNA Clean & Concentrator (Baseclear, Leiden, The Netherlands), and ligated using T4 DNA polymerase (Promega, Madison, WI) to yield pFBTS3-CfTSF1. Escherichia coli DH5α cells were transformed with the ligation mix using a standard heat-shock protocol and transformants were selected in lysogeny broth (LB)-kanamycin agar (50 μg.ml -1 ). Plasmids were extracted from transformants and screened by restriction digest analysis using PacI and NotI in a double digestion. A plasmid bearing the correct restriction pattern was sent to Macrogen (Amsterdam, The Netherlands) for sequencing of the insert. Agrobacterium tumefaciens AGL1 was transformed with pFBTS3-CfTSF1 by electroporation, and plated on LB-kanamycin agar (50 μg.ml -1 ). One positive transformant was picked, verified and named AT-pFBTS3-CfTSF1. This plasmid was introduced to C. fulvum using A. tumefaciens-mediated transformation as previously described [65]. Transformants were selected on PDA medium supplemented with hygromycin (100 μg.ml -1 ). Several transformants and wild-type C. fulvum were grown for 5 days in potato-dextrose broth (PDB; Oxoid, Altrincham, UK) and then transferred to Gamborg B5 medium without nitrogen (B5-N) in order to induce the Avr9 promoter [37,44]. Total RNA was extracted and cDNA synthesis was performed as previously described [65]. The induction of the CfPKS1 biosynthetic gene cluster was confirmed by RT-qrtPCR using primers listed in S1 Table. One transformant showing the expected strong induction of genes at the CfPKS1 locus was selected and named OE.CfTSF1.
Using the same methods, A. tumefaciens AGL1 was transformed with plasmid pRM254, which contains GFP and geneticin-resistance genes [66] to yield AT-pRM254 strain. The plasmid was introduced into the OE.CfTSF1 transformant as described above. Transformants were selected on PDA medium supplemented with geneticin (100 μg.ml -1 ). Transformants were picked and screened for GFP fluorescence. One transformant was selected and named OE. CfTSF1::GFP.

Generation of Δcfpks1 deletion mutants
The plasmid for targeted gene replacement of CfPKS1 was generated following the same procedure as described in Griffiths et al. (2016) [40]. The upstream (US) and downstream (DS) regions flanking of CfPKS1 were amplified using primers 1 + 2 and 3 + 4, respectively (S1 Table), and cloned into pDONR-P4-P1R and pDONR-P2-P3 vectors. The final gene replacement plasmid was assembled in a LR reaction (Invitrogen) that combined the pDONR-P4-P1R::US_CfPKS1, pDONR-P2-P3::DS_CfPKS1, p221_GFP_HYG (pDONR containing a cassette with GFP and HYG resistance marker genes) and the destination vector pDEST R4-R3 [40,66]. One correct sequenced plasmid was chosen and named pDest43-Δcfpks1. This plasmid was introduced into C. fulvum 0WU using the A. tumefaciens transformation method as described in Okmen et al. (2013) [65]. Transformants were selected on PDA plates containing hygromycin (100 μg.mL -1 ). Genomic DNA of each strain was isolated using a Zymo Research Genomic DNA Clean & Concentrator TM (Baseclear), according to the manufacturer's recommendations. PCR and quantitative real-time PCR were performed to screen for double crossovers and measure the number of inserted deletion cassettes, respectively (S1 Fig and S1 Table).

Plant inoculation and determination of fungal growth
Inoculation of tomato with C. fulvum wild-type, deletion mutant and transformant strains was carried out according to a previously described method [67]. To determine fungal growth, the fourth composite leaf of infected tomato plants was harvested at 4, 8, and 12 days post-inoculation (dpi) and flash frozen in liquid nitrogen. Samples were ground to a fine powder in liquid nitrogen, and total RNA was extracted from 100 mg of material using a Zymogen Direct-zol RNA MiniPrep kit (Baseclear) according to the manufacturer's recommended protocol. cDNA synthesis was performed using 100-2,000 ng of total RNA and M-MLV reverse transcriptase (Promega), following the manufacturer's protocol. To assess C. fulvum growth during infection, the actin gene of this fungus was targeted by qrtPCR using the Cf-actin_RT-qrtPCR_F/Cf-actin_RT-qrtPCR_R primer pair method [67]. For sample calibration, the Solanum lycopersicum actin gene was targeted using the Sl-actin_qrtPCR_F/Sl-actin_qrtPCR_R primer pair method [67]. The same cDNA samples were used to measure the expression of genes at the CfPKS1 locus by qrtPCR using previously reported methods and primers [38]. Additional oligonucleotides (S1 Table) were designed and their efficiency determined as described in [38]. Results were analyzed according to the 2 -Δ Ct method [68] and are the average of three biological replicates.

Construction of vectors for heterologous expression and generation of A. oryzae M-2-3 transformants
The cloning of CfPKS1 in heterologous expression vectors was performed as described in Griffiths et al. (2016) [40]. Briefly, CfPKS1 was amplified from C. fulvum OE.CfTSF1 transformant cDNA (grown on B5-N medium) by PCR using primers 5 + 6 (S1 Table) and cloned into NotI-linearized pEYA2 using recombination in S. cerevisiae BMA 64 to generate plasmid pEYA2-CfPKS1 [40,53]. CfPKS1 was transferred into the expression vector pTAex3GS using LR clonase (Invitrogen), and the resulting pTAex3GS-CfPKS1 plasmid was introduced in A. oryzae M-2-3 using PEG-mediated transformation as described in Griffiths et al. (2016) [40]. The starch-inducible taka-amylase promoter (PamyB) controls the expression of CfPKS1. The final vector contains the arginine biosynthesis gene (argB) for selection of fungal transformants.

Secondary metabolite extraction and analysis by LC-MS
Selected A. oryzae transformants containing CfPKS1 were grown on dextrose-peptone-yeast extract (DPY) agar plates at 30˚C until the whole plates were covered. The cultures were freeze-dried and then homogenised with a pestle and mortar. The homogenate was resuspended in water, acidified to pH4 with HCl, and then twice extracted with ethyl acetate. The organic phase was recovered and dried under a nitrogen flow. Samples were resuspended in acetonitrile (CH 3 CN), centrifuged at 20,000 x g for 5 min in a microcentrifuge tube and then transferred to a 1 mL clear glass shell vial (WAT025054c).

Semi-preparative LC-MS, compound purification and structure determination
Purification of compounds was achieved using a Waters mass-directed autopurification system comprising a Waters 2767 autosampler, Waters 2545 pump system, a Phenomenex Kinetex Axia column (5 μm, C18, 100 Å, 21.2 × 250 mm) equipped with a Phenomenex Security Guard precolumn (Luna C5 300 Å) eluted at 20 mL.min -1 at ambient temperature. Solvent A, HPLC grade H 2 O + 0.05% formic acid; Solvent B, HPLC grade CH 3 CN + 0.045% formic acid. The post-column flow was split (100:1) and the minority flow was made up with HPLC grade MeOH + 0.045% formic acid to 1 mL.min -1 for simultaneous analysis by diode array (Waters 2998) and ESI mass spectrometry in positive and negative modes (Waters SQD-2). Detected peaks were collected into glass test tubes. Combined tubes were evaporated under a flow of dry N 2 gas and weighed. HRMS data were measured using Waters Q-Tof Premier operating in ES + mode.