Characterization of NRPS and PKS genes involved in the biosynthesis of SMs in Alternaria dauci including the phytotoxic polyketide aldaulactone

Alternaria dauci is a Dothideomycete fungus, causal agent of carrot leaf blight. As a member of the Alternaria genus, known to produce a lot of secondary metabolite toxins, A. dauci is also supposed to synthetize host specific and non-host specific toxins playing a crucial role in pathogenicity. This study provides the first reviewing of secondary metabolism genetic basis in the Alternaria genus by prediction of 55 different putative core genes. Interestingly, aldaulactone, a phytotoxic benzenediol lactone from A. dauci, was demonstrated as important in pathogenicity and in carrot partial resistance to this fungus. As nothing is known about aldaulactone biosynthesis, bioinformatic analyses on a publicly available A. dauci genome data set that were reassembled, thanks to a transcriptome data set described here, allowed to identify 19 putative secondary metabolism clusters. We exploited phylogeny to pinpoint cluster 8 as a candidate in aldaulactone biosynthesis. This cluster contains AdPKS7 and AdPKS8, homologs with genes encoding a reducing and a non-reducing polyketide synthase. Clusters containing such a pair of PKS genes have been identified in the biosynthesis of resorcylic acid lactones or dihydroxyphenylacetic acid lactones. AdPKS7 and AdPKS8 gene expression patterns correlated with aldaulactone production in different experimental conditions. The present results highly suggest that both genes are responsible for aldaulactone biosynthesis.

www.nature.com/scientificreports/ allowed to identify 5 A. dauci-specific SM genes, underlying A. dauci's potential to produce yet-to-be described SM, including toxins. Our second aim was to better understand the biological activity and biosynthesis pathway of aldaulactone. We first checked aldaulactone leaf toxicity on Nicotiana benthamiana (order: Solanales), a distant species from D. carota (order: Apiales). We then hypothesize that aldaulactone biosynthesis follows the same pattern as for other benzenediol lactones, and involves a HR-PKS and a NR-PKS. Based on our knowledge of RALs and DALs biosynthesis, we identified a candidate cluster for the aldaulactone biosynthesis pathway by homology search, phylogenetic and in silico retro-biosynthesis approaches 52 . Furthermore, we pointed out a significant correlation between aldaulactone production and the expression level of the HR-PKS (AdPKS7) and the NR-PKS (AdPKS8) genes belonging to this candidate cluster. Our data indicates that the cluster containing AdPKS7 and AdPKS8 may be responsible for aldaulactone biosynthesis. Moreover, our results further validate the potential to discover novel toxins involved in both A. dauci pathogenicity and D. carota partial resistance.

Results
Aldaulactone phytotoxicity test on tobacco. In our previous paper 39 , only in vitro proofs were provided for aldaulactone toxicity. Direct in planta evidence of aldaulactone toxicity was not yet established, since carrot leaf infiltration with the toxin is challenging, because of leaf fragility (results not shown). Alternatively, experiments of plant infection or infiltration were performed using the N. benthamiana model. Necrotic leaf lesions (black spots surrounded by a yellow halo) on tobacco were obtained after inoculation with a conidial suspension of FRA001 strain (Fig. 1a). Isolation from symptomatic leaf pieces allowed us to obtain fungal colonies producing conidia exhibiting a typical A. dauci morphology. From those isolates, species identification was done by sequencing portions of three target genes (ITS, EF1-α, IGS). The obtained sequences exactly matched those of A. dauci strain FRA001 (data not shown). These results showed that A. dauci is pathogenic on tobacco in our experimental conditions.
Meanwhile, phytotoxicity tests were performed by leaf infiltration using two aldaulactone concentrations and an organic extract of ITA002 culture medium (Fig. 1b,c). Lesion areas nine days post-infiltration of the control conditions (0.1% DMSO and PDB) were in the same statistic class than the ones observed after infiltration using 12.5 µg mL −1 aldaulactone. The lesion areas produced by the infiltration with 50 µg mL −1 aldaulactone or by the ITA002 organic extract were significantly higher than control conditions. Aldaulactone is able to produce necrotic lesions on tobacco leaves. Improvement in Alternaria dauci genome assembly by RNA-seq-mediated and reference genome scaffolding.. The assembly of the only A. dauci genome published on a database is composed of many small contigs (N50 = 13,282 bp; number of contigs = 12,030), shorter than the typical length of SM gene clusters, that amount to several tens of kilobases 55 (Table 1). Moreover, A. dauci genome assembly contained only 72.2% Pezizomycotina BUSCOs genes with 14.5% fragmented and 13.2% missing data (Fig. 2). Two complementary scaffolding strategies were performed to improve the quality of the assembly: (1) genome assembly improvement using the AGOUTI tool 56 with the RNA sequencing data from FRA001 strain (available at http:// www. ncbi. nlm. nih. gov/ biopr oject/ 790446), and (2) reference-based (A. solani genome 57 ) genome re-assembly with the CSAR tool 58 . After applying the first strategy, the number of contigs/scaffolds (> 100 bp) decreased from 4,010 to 3,139 and N50 increased from 13,282 to 18,857 bp (Table 1). Due to the method, the assembly improvement was expected only in the expressed areas of the genome, explaining the apparently modest gains made. Nevertheless, many contigs containing SM cluster were reassembled.
The second scaffolding strategy gave a new improved assembly called below "CSAR genome assembly". When applying the second scaffolding strategy, a strong improvement in genome contiguity was obtained with a decrease from 3139 to 553 scaffolds (> 100 bp) and a N50 increase from 18,857 bp to 4.49 Mpb (Table 1). The BUSCO completeness of 78.1% with 9.6% fragmented and 12.2% missing data was largely improved over the 72.2% of the published genome assembly even if the percentage remain still relatively low (Fig. 2). This scaffolding therefore improved genome contiguity and gene completeness, and produced an assembly including 7594 scaffolds with a total genome size of 33.4 Mb and a GC content of 50.68% (Table 1).
The RNA-Seq library was constructed from FRA001 strain and sequenced using Illumina HiSeq2000: 161.96 million of paired-end reads and 160.5 million of cleaned reads were obtained. Reads mapping on CSAR genome assembly was performed by HISAT2. A total of 87.8% of those reads aligned concordantly one time. After this step, StringTie and Cufflinks were used to assemble the RNA-Seq alignments into 12,939 transcripts ( Table 1). The unigene N50 was 1713 bp consistent for a fungal transcriptome.  Fig. 4a). Sequences 11 and 13 were only predicted in the first genome annotation. Sequences 6 and 13 were consistent with a single domain. We hypothesized that these 2 sequences do not correspond to PKS genes. For sequence 3 (KS-AT-DH) and 11 (AT-DH-ER-KR), misassembly could cause the missing domains or these two sequences may correspond to pseudogenes. Sequence 19 (SAT-partial KS) was localized in cluster 8 in scaffold 3 and took end in a sequencing gap of the genome. Sequence 20, encoding a NR-PKS 3'-end (AT-PT-ACP-TE), was consistent with the whole scaffold 116 and was found to be expressed. Primers designed in sequence 20 and on both sides of the scaffold 3 To study the putative functions of those predicted proteins, homologies with characterized proteins were pointed out (Supplementary Tables S1, S2). A. alternata TES (A0A144KPJ6.1) and TES1 (A0A144KPK9.1), two enzymes responsible for tentoxin biosynthesis, were found to have more than 80% identity with AdNRPS1 and a protein encoded by another cluster 1 gene. AdNRPS2 and AdNPRS3 are linked to siderophore-mediated iron metabolism. Indeed, AdNRPS2 in cluster 9 showed 59% identity with Bipolaris maydis intracellular siderophore synthetase (Q5D6D7.2), whileAdNRPS3 in cluster 1 exhibited 86% similarity with Bipolaris oryzae extracellular siderophore synthase (Q09MP5.1) 59 . AdPKS6 in cluster 7 showed 95% similarity with A. alternata melanin synthase (BAK64048.1). As previously described 44 , A. dauci genome contains a homologue of A. solani alternariol synthase cluster (cluster 2). AdPKS12 in cluster 13 showed 91% identity with A. solani alternapyrone synthase (Q5KTM9.1) and the whole cluster was conserved. AdPKS13 in cluster 14 had 91% identity with A. solani aslaniol synthase (Q2ABP6.1). At last, AdPKS14 in cluster 15 presented 95% identity with Alternaria cinerariae Dhc5 (dehydrocurvularin biosynthesis protein 5, KT271474.1). No homologue to Dhc3, the other PKS necessary for dehydrocurvularin biosynthesis (KT271472.1) was found in the A. dauci genome.  Table 1. Assembly summary statistics applied to the different assemblies of Alternaria dauci genome data set. AGOUTI re-assembly: re-assembly by the AGOUTI tool 56 of the published genome 51 helped by FRA001 transcriptome data obtained in this study. CSAR scaffolding: reference-based (A. solani genome 57 ) genome re-assembly with the CSAR tool 58 . N50: length of the shortest contig/scaffold in the smallest subset of contigs whose length sum makes up half of genome size. L50 number of contigs in that subset. GC (%): guaninecytosine content.

Assembly
Published genome 55 AGOUTI re-assembly CSAR scaffolding  PKS and hybrid core genes within each species. Blue box: gene is present. Genes present in the same column share more than 80% identity and are considered orthologues. Gray triangle: a partial copy or a pseudogene is present. Black box: gene was not found. (c) UPGMA dendrogram generated from a binary matrix of SM core gene presence and absence derived from (b) with the use of Jaccard coefficients to compare between sets of variables. This dendrograms follows closely the phylogenetic tree shown in (a) except for four species only (orange streaks). www.nature.com/scientificreports/ those of the alternata section and even more than A. brassicicola. To support this observation, a dendrogram was produced from a gene presence/absence matrix and compared with the phylogeny of the studied Alternaria strains (Fig. 5c). This allowed to differentiate most species, excepting A. tangelonis and A. tenuissima, while this phylogenetic analysis also failed to distinguish A. tenuissima and A. alternata. Both dendrograms showed a similar strain repartition in three clades: (1) the porri section strains, (2) the alternata section strains, and (3) A. brassicicola. Moreover, strain distribution within each clade was strongly similar in both dendrograms.

Phylogeny of KS and PT domain of A. dauci PKS. A phylogenetic study of the KS and PT domains
was performed to classify the AdPKS sequences among characterized PKSs from Dothideomycetes and to find candidates for aldaulactone biosynthesis, i.e. PKSs homologous to NR-PKS and HR-PKS involved in DAL biosynthesis. Eighty-two Dothideomycete KS-sequences from 60 and from the UniProt reviewed database were used for a maximum of parsimony tree reconstruction (Fig. 6, Supplementary Table S3). The clades produced and AdPKS-domain structures predicted were consistent to those previously described 60  Interestingly, HR-PKS clade I contained a very well defined subclade (bootstrap value of 99%) that regrouped AdPKS7 and all HR-PKS involved in benzenediol lactone biosynthesis. AdPKS4, AdPKS12 and AdPKS7, clearly belonged within HR-PKS clade I, but not in the same subclade as AdPKS7. AdPKS2, AdPKS5 and sequence 3 clustered in HR-PKS clade II. AdPKS9 clustered within the HR-PKS clade IV. AdPKS15 and the sequence 6 clustered in the PR-PKSs clade. The NR-PKS clade basal to clade I&II contained two monophyletic subclades, one including AdPKS1 and AdPKS10, the other gathering AdPKS8, AdPKS14 and NR-PKSs involved in benzenediol lactone biosynthesis. NR-PKS clade II included AdPKS6. AdPKS3 and AdPKS11 clustered in NR-PKS in clade III. According to KS-domain phylogeny, AdPKS7 and AdPKS8 or AdPKS14 were consistent candidates for aldaulactone biosynthesis.
The phylogenetic tree of PT domains was constructed from 35 NR-PKSs, including 6 AdPKSs, to study firstring aldol-cyclization stereoselectivity (Fig. 7). PKSs were grouped in 5 monophyletic clades consistent to those described by 27 . AdPKS6 PT domain clustered within clade II. AdPKS1 and AdPKS10 clustered within clade V, respectively with C2-C7 and C6-C11 type PT domains. Clade I contained two monophyletic subclades, the first one clustered C2-C7 type PT domains, the second one C3-C8 type PT domains. AdPKS11 belongs to the first subclade, AdPKS14 and AdPKS8 to the second one. These results strengthened the fact that AdPKS14 and AdPKS8 are good candidates for aldaulactone biosynthesis.
Functional analysis of AdPKS for aldaulactone production. In order to decipher aldaulactone biosynthetic pathway, the correlation between expression levels of 8 AdPKS genes and aldaulactone accumulation was investigated by HPLC-DAD and RT-qPCR experiments (Fig. 8). To produce variations in aldaulactone production and AdPKS genes expression, the experimental conditions consisted of (1) 4 A. dauci strains grown in PDB medium, and (2) A. dauci FRA001 strain grown in different conditions. The FRA001 strain grown in PDB medium was taken as a reference and all results are expressed as ratios.
In the same trends as in 39 , FRA017 and FRA001 strains had a same order of aldaulactone production, while AUS001 produced about 30-fold less and ITA002 strain produced between twice and thrice more aldaulactone than FRA001 strain. Considering FRA001 strain, a non-statistically significant higher production of aldaulactone was found in minimal medium (Vogel) and in PDB with zebularine, by comparison with PDB. The addition of DMSO in PDB significantly decreased aldaulactone production, while addition of SAHA or trichostatin had no significant effect (when compared with DMSO). Aldaulactone concentration was below the limit of detection when FRA001 was grown in PDB, without medium agitation during 21 days.
The expression ratios of the 8 AdPKS genes in the different experimental conditions presented various patterns. AdPKS10 expression was repressed in ITA002 strain compared to FRA001, FRA017 and AUS001 strains. Interestingly, for the other AdPKS genes studied, the opposite pattern was observed: transcription levels seemed to increase with strain aggressiveness, AUS001 being the less aggressive and ITA002 the more aggressive as reported in 61 .
The Spearman correlation coefficient between ratios of AdPKS expression and aldaulactone yields were calculated and 5 of them were statistically significant (p < 0.01; Fig. 8c). Among them, the one obtained for AdPKS10 was negative, while others were positive. The correlation coefficient obtained for AdPKS14 (r = 0.46) was almost twice weaker than those obtained for AdPKS-NRPS5, AdPKS7 and AdPKS8 genes (r ≥ 0.81). Furthermore, AdPKS-NRPS5, AdPKS7 and AdPKS8 expressions were cross correlated.

Discussion
A. dauci belongs to the Dothideomycetes, which are known to produce various SM. More than 250 SM have been described from Alternaria genus members, a lot of them being HST or NHST 1,62 . Interestingly, strong evidence of toxins' role in A. dauci aggressiveness and D. carota partial resistance were provided 37 . Among the few toxins investigated in A. dauci (zinniol, alternariol, alternariol monomethyl ether…) [39][40][41][42][43] , aldaulactone was supposed to explain most, but not all, in vitro toxicity of A. dauci exudates. However, in planta evidence of aldaulactone toxicity was compromised due to the experimental barrier when using carrot leaves 39  www.nature.com/scientificreports/ the pathogenicity of A. dauci and the aldaulactone toxicity on tobacco leaves. Despite the importance of toxins in the A. dauci pathogenicity, the current understanding of its SM production and relevant biosynthetic pathways is still very limited. We provided the first transcriptome of A. dauci and the first investigation of SM core gene diversity within Alternaria genomes, including A. dauci. Finally, we predicted the aldaulactone biosynthesis cluster by a comprehensive phylogenetic analysis and a correlation between aldaulactone production and expression of PKS genes. From the structure of both the toxin biosynthesis cluster and aldaulactone, we proposed a biosynthetic pathway for this toxin (Fig. 9). The transcriptomic data of A. dauci were used, to substantially improve the A. dauci genome assembly completeness. A. dauci's genome contained 19 predicted SM clusters, comprising 15 PKS genes, 1 hybrid PKS-NRPS gene, 1 PKS-like gene and 6 NRPS genes. When this bioinformatic prediction in A. dauci was compared to the number of PKS gene clusters predicted from 75 Dothideomycetes, A. dauci was positioned at the 29th rank 63 . Thus, the number of PKS genes is relatively high, by comparison with a phylogenetically distant Alternaria species like A. brassicicola (situated at the 62nd rank).
From genomic data available online, we predicted, 55 groups of putative homologous SMs core genes in 21 strains belonging to 19 Alternaria species. To our knowledge, no genomic comparative study focused on SMs genes have already been performed at the Alternaria genus scale. Moreover, few SM biosynthetic pathways were characterized within Alternaria, despite numerous studies using metabolomic profiles and the importance of SMs in the lifestyle of those necrotrophic fungi 39,41,42,62 . Each Alternaria genome harbored a different set of SM core genes as shown in Fig. 5, depending on species phylogenic classification. Remarkably, among the 34 PKSencoding genes, only 7 NR-PKSs were predicted and all were found in the A. dauci genome. The dendrograms www.nature.com/scientificreports/ either based on SM core gene binary repartition or on Alternaria spp. phylogenic relationships gave the same three separated clades: strains belonging to the porri section (1), or to the alternata section (2), and the sole A. brassicicola strain (3). Alternaria strains inside the porri section were also grouped in the same manner by SM core gene criterion and molecular taxonomy (Fig. 5). Contrastingly, Alternaria strains in the alternata section were not similarly segregated when comparing both analysis tools. Here, patterns of SM core gene were used to help in Alternaria taxonomy, as chemotaxonomy analysis. In fact chemotaxonomy was demonstrated to be efficient in Alternaria strains/species discrimination in the porri section 42,64 . Our SM core gene prediction could be considered as an illustration of a potential secondary metabolome in Alternaria spp. independently of culture conditions. In further analyses, the comparison between the presence/absence of SM core genes in Alternaria strains and their corresponding metabolic profiles would be useful to identify new biosynthetic pathways. Among the 55 SM core gene types, 36 had weak similarities to characterized SM genes. This suggests that the relevant metabolites produced may not yet be linked to known biosynthetic pathways or may correspond to unidentified compounds. The remaining 19 SM core gene types presented homologies with genes involved in melanin production and core genes known to be involved in the biosynthesis of pathogenicity factors (Supplementary Table S2) 65 . Twelve and four of these last types are respectively involved in biosynthetic pathways of NHST and HST (AF-, ACR-, AAL-and AM-toxin). Other SM core genes, involved in ferricrocin (intracellular siderophore) and extracellular siderophores, may indirectly play a role in fungal pathogenicity by improving iron uptake 59 . www.nature.com/scientificreports/ Interestingly, the SM core genes present in all studied genomes encoded NHST and those present in a single genome encoded HST. Eighteen genes were private to a single Alternaria genome, highly suggesting the production of specific SM in the relevant strains. Also, A. dauci genome contained genes responsible for the biosynthesis of already described compounds: alternariol, melanin, alternapyrone, aslaniol, tentoxin, ferricrocin, extracellular siderophore, 6-methylsalicylic acid and one of the two genes involved in dehydrocurvularin biosynthetic pathway. Moreover, five of the A. dauci genes described here (AdPKS2, AdPKS7, AdPKS8, AdPKS16 and AdNRPS6) were private. Among them, AdPKS7 and AdPKS8, were identified as candidate genes for aldaulactone biosynthesis. Here, transcriptomic data showed expression of AdPKS1, AdPKS3, AdPKS6, AdPKS7, AdPKS8, AdPKS10, AdPKS13, AdPKS16, AdNRPS1, AdNRPS2 and AdNRPS3, highly suggesting the potential ability for A. dauci to produce melanin, alternariol, aslaniol, and tentoxin. This would be in line with the fact that aldaulactone was not the unique source of toxicity in A. dauci exudates as developed in 39 .
The production of alternariol, alternariol monomethyl ether, zinniol and aldaulactone by A. dauci was demonstrated from in vitro cultures 40,42,43 . The tentoxin biosynthesis core gene was present in A. dauci, but was not expressed in our culture conditions, nor in previous studies 42, 66 . This could be linked to specific culture conditions which did not allow gene expression and SM production, as previously observed for zinniol: in vitro zinniol production is drastically controlled by A. dauci culture conditions 40 and occurs only in long lasting cultures without oxygenation and not in 48 h oxygenated cultures 37 . In order to better understand pathogenicity mechanisms, it would be interesting to decipher the expression patterns of SM core genes during the carrot leaf infection by A. dauci. Evidence of other involved pathogenicity factors, like cutinolytic enzyme activity was provided by the observation of subcuticular hyphae 67 . The occurrence of A. dauci lesions on carrot and the fungal host range could be partly explained by a joint action of lytic enzymes and a toxin cocktail mixing NHST and HST, as usually observed for necrotrophic fungi.
Concerning A. dauci host range, a previous study showed that the fungus was able to produce symptoms and sporulate mainly on different Apiaceae species and secondarily on cultivated species belonging to Brassicaceae, Solanaceae and Valerianaceae 61 . In the present paper, necrotic lesions provoked by A. dauci and a toxic effect of aldaulactone, at concentrations found in fungal in vitro cultures were observed on N. benthamina leaves. Aldaulactone was thus responsible for necrotic symptoms in a Solanaceae species, which did not belong to the main A. dauci host range. This is in contradiction with the idea that aldaulactone could be classified as an HST.
In order to decipher aldaulactone biosynthetic pathway, a retro-biosynthesis approach was used, based on our SM core gene predictions and on the DAL nature of aldaulactone. While several RAL biosynthetic pathways are known among fungi, only one DAL (dehydrocurvularin) biosynthetic pathway was fully described, within two fungal species 47 . Through KS-domain phylogenetic analysis, we highlighted that enzymes responsible for DAL and RAL biosynthesis clustered into two subclades of NR-PKS clade basal to clade I and II and HR-PKS clade I (Fig. 6). Three A. dauci candidate genes-AdPKS7, AdPKS8 and AdPKS14-clustered within these two subclades. Furthermore, our phylogenetic approach on NR-PKS PT domains showed that AdPKS14 and AdPKS8 PT domains clustered with PT domains catalyzing C3-C8 cyclisation.
To assess the candidate genes involved in aldaulactone synthesis, qPCR and HPLC experiments were conducted. Liquid cultures obtained in different conditions led to a wide range of gene expression and aldaulactone production. Different studies have shown an epigenetic regulation of fungal SM gene expression. In particular, drugs affecting epigenetic regulation are used to induce the expression of fungal genes, which are weakly or nonexpressed under in vitro conditions [68][69][70][71] . In the present study, two histone deacetylase inhibitors, trichostatin A www.nature.com/scientificreports/ and suberoylanilide hydroxamic acid (SAHA) and a DNA methylation inhibitor (zebularine) were tested, but no significant induction effect on gene expression was observed. These results might be due to the too short exposition time of fungal cultures with the tested drugs 71 . Among A. dauci PKS genes, a significant correlation between the expression level of four genes (AdPKS-NRPS5, AdPKS7, AdPKS8 and AdPKS14) and aldaulactone production was highlighted. Because of its hybrid nature, AdPKS-NRPS5 was not further investigated as a candidate for aldaulactone biosynthesis. Here, a very low expression of AdPKS14, was observed under growth conditions suitable for aldaulactone production. Furthermore, AdPKS14 did not cluster with HR-PKS genes and its expression was not correlated with expression of other PKS genes. Contrastingly, AdPKS7 and AdPKS8 are good candidates for the biosynthesis of aldaulactone carbon backbone, as both genes belong to the same cluster (cluster 8; Fig. 3). The fact that expression of AdPKS-NRPS5, AdPKS7 and AdPKS8 is cross correlated, could be interpreted as a co-regulation of their expression associated to the fungal aggressiveness.
According to all results presented in this study, a biosynthetic pathway based on cluster 8 was hypothesized (Fig. 9): AdPKS7 catalyzed the biosynthesis of a reduced aliphatic PK, which was then used by AdPKS8 as a precursor. AdPKS8 extended that PK, and then its PT domain would catalyze a C3-C8 cyclization leading to the benzenediol backbone. At last, the TE domain released the molecule by a ten-membered macrolactone cyclization. Cluster 8 also contained 12 other genes with or without predicted functions. In particular, cluster 8 contained the enzymes required for the addition of a methoxy group on C6: an oxygenase-encoding gene ( Fig. 3; rightmost ORF in cluster 8) and a methyl transferase-encoding gene (eleventh ORF from the left). Both enzymes were found to be expressed in transcriptomic data. The oxygenase would add a phenol function to the benzenediol moiety and finally the methyl-transferase gene would methylate this function. The others genes in cluster 8 were also found to be expressed except for the fourth and the sixth ORFs from the left. Those expressed genes include two putative transporter genes, one of them was predicted as a multidrug resistance transporter by fungismash prediction. To our knowledge, aldaulactone was the second example, after dehydrocurvularin 47,72 , of C3-C8 cyclization catalyzed by a PT domain in fungi.
An original integrative approach of retrobiosynthesis combining phylogenetic analysis, gene expression study, and HPLC quantification was used to decipher aldaulactone biosynthetic pathway. This work not only presented the first transcriptomic analysis in A. dauci, but also provided the first study of diversity and repartition of SM genes through Alternaria genus. Moreover, it revealed the putative biosynthetic pathway of aldaulactone, a DAL type phytotoxin, whose role is crucial in A. dauci pathogenicity.

Methods
Fungal material. The same four A. dauci strains and one A. brassicicola strain as in 39 61 . Numerous studies were also conducted on strain FRA001 37,39,61,67 . Those strains were collected as described in 61,73  Fungal growth conditions for transcriptomic samples. FRA001 conidial suspensions for transcriptome were prepared as follows. Three 5 mm mycelial plugs of strain FRA001, previously cultivated on Malt-Agar medium as described in 67 , were placed on a sterile cellophane membrane in a Petri dish (90 mm diameter) containing V8 ® agar medium [175 mL of vegetable juice V8 ® (Campbell Soup Company), 3 g of calcium carbonate (CaCO 3 ), 15 g of bacteriological agar, final volume adjusted to 1 L with ultrapure water, pH 6.8]. Fungal cultures were incubated in the darkness at 20 ± 2 °C for 10 days. The conidial suspension was prepared by adding 6 mL of 0.1% Tween 20, scraping it with a sterile glass rake, then filtering through two layers of gauze. Conidial density was evaluated on a Malassez cell and adjusted to the required concentration.
Three different culture methods were performed. For the first culture method, a sterile cellophane membrane placed on the surface of a Petri dish (90 mm in diameter) containing "carrot juice" agar medium (200 mL of 100% pure carrot juice, Eckes-Granini, Joker ® ), 3 g calcium carbonate (CaCO 3 ), 15 g bacteriological agar, final volume adjusted to 1 L with ultrapure water, pH 6.8) was inoculated with 2 mL of conidial suspension containing 2 × 10 5 conidia. This culture was then incubated for 24 h at 22 °C in the darkness. For the second and third culture methods, 5 × 10 6 conidia were inoculated in 100 mL of V8 ® liquid medium placed in a 250 mL Erlenmeyer flask and incubated for 24 h at 22 °C in the darkness. One culture was maintained with 125 rpm checking, the other one without shaking. Whatever the culture method, the collected germinated conidia were immediately immersed into liquid nitrogen and then stored at − 80 °C until RNA extraction. www.nature.com/scientificreports/ 15,000 g and 4 °C and the supernatant was discarded. The pellet was washed with 500 µL of 75% ethanol. A centrifugation at 15,000 g and 4 °C for 30 min was performed to remove ethanol. The pellet was dried and then dissolved in 30 µL of "RNAse-free" ultrapure water. Evaluation of RNA integrity and concentration was performed by the Experion ® automatic electrophoresis system generating a RNA quality indicator (RQI) and concentration data. RNA with a RQI between 7 and 8 were selected.
Illumina sequencing and de novo assembly. Preparation of RNA-seq library from a pool of fungal RNA, from the tree cultural methods described above, RNAseq protocol and de novo assembly were performed by NGS Services Fasteris (Plan-les-Ouates, Switzerland). cDNA sequencing was conducted on an Illumina HiSeq2000 platform using 100 bp paired-end sequencing strategy. Adapter sequences were removed. The reads were then assembled de novo using the Velvet (1.2.07) software with the additional module Oases 75,76 . A first functional annotation was performed by research of unigenes by clustering the top-hit from BLASTX searches in NR database of NCBI 77 . An InterProScan analysis with Hmmpfam, blastProDom, FPrintScan, ProfileScan applications was performed using Blast2GO 78 .

Improvement of A. dauci genome assembly.
The publicly available genome of A. dauci 55 contig datas were used for further scaffolding. AGOUTI was used to stitch genome contigs together based on BWA mediated alignment of transcriptome paired short reads on genome. Ns were added when sequences are unknown 56,79 . A second scaffolding was performed using the CSAR online tool with the "NUCmer on nucleotides" option and A. solani genome as reference 57 . A structural annotation was processed by the AUGUSTUS online tool 80 on the new genome assembly. The gene prediction was done on both strands allowing few alternative transcripts and based on Botrytis cinerea as model organism. GenomeQC was used to check assembly contamination and evaluate the metrics and completeness with Pezizomycotina BUSCO dataset 81 .

PCR and sequencing of gaps.
To manage gene and cluster reconstruction, PCR amplification of gaps between two sequences was performed. For that purpose, genomic DNA from FRA001 mycelium was extracted according to 82 . Primers used are described in Supplementary Table S4 SM core gene and cluster prediction. Putative PKS and NRPS gene clusters were predicted from transcriptomic and genomic data using computational tools specialized in gene identification of fungal SM, such as antiSMASH (fungiSMASH version) 86,87 and SMURF 88 . AntiSMASH (fungiSMASH version) prediction was performed on the 21 Alternaria genomes. The PKS and NRPS domains were predicted from the deduced unigene protein using the NCBI Conserved Domain Search and the Hidden Markov Models obtained from PFAM 89 . Presence or absence of each gene was checked in all analyzed genomes by tBlastN research. If the sequences showed more than 80% of similarity in 80% of their length, they were put in the same set of genes. For A. dauci genome CSAR assembly, SMURF was also used to predict SM clusters with AUGUSTUS structural annotation. When minimal domains were detected, i.e. A-C-PCP for NRPS and KS-AT-ACP for PKS, the relevant genes were renamed using "AdNRPS" or "AdPKS" as a prefix and a number. This numbering followed the same order than the sequence numbers in the genome. www.nature.com/scientificreports/ An Alternaria phylogenetic tree was generated from sequences of four housekeeping genes-Alternaria major allergen gene (Alt a 1), glyceraldehyde-3-phosphate dehydrogenase (gapdh), translation elongation factor 1-alpha (tef1) and RNA polymerase second largest subunit (rpb2)-found in the set of the 21 Alternaria genomes mentioned above. The sequences of each gene were aligned, manually adjusted and then concatenated. The alignment of concatenated sequences was generated with MAFFT v. 7 (http:// mafft. cbrc. jp/ align ment/ server/ index. html). Findmodel (http:// www. hiv. lanl. gov/ conte nt/ seque nce/ findm odel/ findm odel. html) was used to choose the nucleotide substitution model. Bayesian analyses were performed with MrBayes v. 3.2.6 (http:// www. phylo geny. fr/ one_ task. cgi? task_ type= mrbay es) on the concatenated sequences aligned dataset. A GTR model with gamma-distributed rate variation was used and a Markov Chain Monte Carlo analysis was performed with 10,000 generations from a random tree topology. RAxML v. 0.9.0 was additionally run on the concatenated sequences aligned dataset to performed a maximum-likelihood analysis including 1000 bootstrap replicates.

Expression patterns of A. dauci putative PKS genes by RT-qPCR and HPLC analysis. Liquid
cultures of A. dauci were obtained by inoculation of 100 mL of medium with 3 mycelial agar plugs in 250 mL Erlenmeyer flasks and growth at 24 °C in the darkness as previously described 39 . Two kinds of experiments were performed. On the one hand, different strains (AUS001, FRA001, FRA017 and ITA002) were grown in stirred Potato Dextrose Broth (PDB, 24 g L −1 ) medium at 125 rpm for 60 h. On the other hand, FRA001 strain was grown under different conditions as described in Table 2. Two different media were used, PDB (24 g L −1 ) and Vogel minimal medium 93 . All cultures were conducted in three replicates separated in time. Briefly, after incubation, culture medium and mycelium were collected by filtration 39 . The mycelium was stored at − 80 °C until RNA extraction. Organic extracts were obtained from filtered culture medium by liquid-liquid extraction with ethyl acetate and dried 37,39 . Absolute quantification of aldaulactone by HPLC was performed on organic extracts from culture filtrates as described in 39 .
Total RNA was isolated from mycelium by grinding in a mortar with the lysis buffer according to the manufacturer instructions (NucleoSpin ® RNA Plus kit,Macherey-Nagel). RNAs were further treated with the Turbo DNA-free kit (Ambion ® , ThermoFisher scientific). Quality assessment was performed using a NanoDrop spectrophotometer. When needed, RNA purification was conducted as previously described above. Reverse transcription (RT) was performed on 1 µg of RNA diluted in a final volume of 9 µL RNase-free ultrapure water and heated for 3 min at 80 °C. The RT reaction was performed on heated RNA in a total volume of 30 µL containing 10 pmol oligo dT(15), 0.1 µg of random hexamer, 1 × RT buffer, 0.5 mM dNTP, 200 U of M-MLV Reverse Transcriptase (Promega). The mix was incubated for one hour at 37 °C and finally 10 min at 80 °C. cDNAs were diluted tenfold with RNase free water and stored at − 20 °C until use.
Primers for PKS genes and housekeeping genes (tef1, Alt a 1, and gapdh) were designed using Perlprimer 94 based on the predicted gene sequences (Supplementary Table S4). All qPCR reactions were conducted from 2 µL of cDNA obtained as described above in a reaction volume of 10 μL containing 1× master mix (Promega) and 0.1 µM of each primer in RNase-free ultrapure water. Melting curves were checked to assess the specificity of primers. An equimolar pool of cDNAs was realized and four ten-fold dilutions (10 to 10 4 ) of this pool were then used as standards. To determined primer efficiency, real-time PCR reactions were performed on standards using StepOnePlus™ Real-Time (RT) qPCR System (Applied Biosystems). When needed, for genes expressed at very low level, cDNA fragments were amplified by PCR using the GoTaq ® Flexi DNA Polymerase PCR kit (Promega) in a 25 μL reaction volume (1 µL of diluted cDNA, 1× GoTaq ® Flexi Buffer, 1.5 mmol L -1 MgCl 2 , 0.2 mM of each dNTP, 0.2 mM of each primer and 0.5 U of GoTaq ® DNA Polymerase). PCR reactions were conducted with the following parameters: 2 min at 95 °C, followed by 30 cycles (95 °C, 1 min; 60 °C, 15 s; 72 °C, 30 s) and finally 2 min at 72 °C. The PCR amplifications were then purified using a Nucleospin gel and PCR clean-up kit according to the manufacturer's protocol (Macherey-Nagel). The purified cDNAs were eluted in a final volume of 15 μL of elution buffer (5 mM Tris-HCl, pH 8.0). The cDNAs concentration in ng μL −1 was measured using a Nanodrop spectrophotometer, then the copy number of cDNAs per microliter was calculated. A final concentration of 0.5 × 10 5 cDNA copies µL −1 was added to cDNA pool before dilutions in ten-fold series. qPCR experiments were performed on 384 wells microplate with each experimental condition tested in triplicates. For each primer mix, a negative control (water), a positive control (genomic DNA of A. dauci) and serial cDNA dilutions as standards were used in the same plate than all the analyzed samples. The deposition of cDNA and reaction mixture (primer and master mix) in the microplates was carried out using a Zephyr Compact Liquid Handling Workstation (CaliperLife Science) robot controlled by the Caliper Life Maestro Workstation Software. www.nature.com/scientificreports/ The qPCR plates were monitored by Biorad CFX384 machine. For each AdPKS gene in each sample, the Ct value was compared with a mean of the Ct values obtained from the three reference genes. A method based on 95 was used to calculate expression ratios. All data were compared using a Kruskal-Wallis test.

Assessment of A. dauci pathogenicity and aldaulactone toxicity on tobacco leaves. Model
plant N. benthamiana seeds were obtained in compliance with relevant institutional, national, and international guidelines and legislation. In the greenhouse, N. benthamiana seedlings were transplanted 15 days after sowing in 75 centiliters pots containing potting soil (Substrate 5, Klasmann ® ) and axillary branches were removed. Plants were maintained with a 16-h photoperiod, a day/night temperature of 23/19 °C and a relative humidity of about 70%. The three first true developed leaves of six-week-old plants were used for two experimental conditions consisting of leaf inoculation or leaf infiltration. First, a conidial suspension of FRA001 strain was prepared according to 61 with a final concentration of 1000 conidia mL −1 . For plant infection, the conidial suspension was spread with a sterile brush on the abaxial leaf side. A mock inoculation was also performed. Second, the organic phase obtained from filtrates of ITA002 liquid culture (100 mL) or PDB medium (100 mL, negative control) was used for leaf infiltration. The filtrates were obtained as described in 39 and dissolved in DMSO to obtain a maximal concentration of 0.1% in a final volume of 100 mL (volume adjusted with ultrapure water). As a reference, a 0.1% DMSO solution (negative control) and two aldaulactone solutions at final concentrations of 12.5 mg L −1 and 50 mg L −1 were prepared. The different filtrates or solutions were infiltrated at five points on a same leaf by pressing a 1 mL needleless syringe on the abaxial leaf side. Two biological replicates were performed. Nine days after infiltration, the leaves were collected, scanned at a resolution of 300 dpi using an Epson Perfection 3200 Pro flatbed image scanner. Images obtained were analyzed using the FIJI software 96 to quantify necrosis area. For inoculated plants with FRA001, isolation from symptomatic leaf pieces was realized on PDA medium (amended with streptomycin 500 mg L −1 ) and incubated as described in 61 . DNA was extracted from mycelial colonies and from FRA001 following the protocol described in 82 . PCR experiments were conducted on three targeted sequences: ITS, tef1 and IGS (primer sequences in Supplementary Table S4). The amplification products were purified and sequenced by Eurofins Genomics. Sequences were manually reassembled and then aligned with relevant sequences obtained from FRA001. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.