Chromosome level assembly and secondary metabolite potential of the parasitic fungus Cordyceps militaris

Cordyceps militaris is an insect pathogenic fungus that is prized for its use in traditional medicine. This and other entomopathogenic fungi are understudied sources for the discovery of new bioactive molecules. In this study, PacBio SMRT long read sequencing technology was used to sequence the genome of C. militaris with a focus on the genetic potential for secondary metabolite production in the genome assembly of this fungus. This is first chromosome level assembly of a species in the Cordyceps genera. In this seven chromosome assembly of 33.6 Mba there were 9371 genes identified. Cordyceps militaris was determined to have the MAT 1-1-1 and MAT 1-1-2 mating type genes. Secondary metabolite analysis revealed the potential for at least 36 distinct metabolites from a variety of classes. Three of these gene clusters had homology with clusters producing desmethylbassianin, equisetin and emericellamide that had been studied in other fungi. Our assembly and analysis has revealed that C. militaris has a wealth of gene clusters for secondary metabolite production distributed among seven chromosomes. The identification of these gene clusters will facilitate the future study and identification of the secondary metabolites produced by this entomopathogenic fungus.


Background
Entomopathogenic fungi are a fascinating group of insect parasitic microbes, which include species from a variety of different fungal taxa including Beauvaria, Hirsutella, Metarhizium Cordyceps and Ophiocordyceps (Fig. 1). These entomopathogenic fungi typically have a lifecycle in which the host is infected and is killed in the process of fungal propagation. Two closely related genera of entomopathogemic fungi, Cordyceps and Ophiocordyceps (often times just referred to as cordyceps in common literature) are characterized by their unique lifecycle and a specific process by which they parasitize and reproduce using the insect host. Fungal spores and hyphae are able to penetrate insect cuticle and then colonize and proliferate within their body cavity. As the insect's body is used as a nutrient reservoir for growth, the insect's behaviour is modified, eventually leading the host to die in an advantageous location for fungal spore dispersal. The fungus then emerges as a fruiting body from the corpse of the insect, which matures and disperses spores of the next generation. Spread globally, Cordyceps and Ophiocordyceps fungi have been described in climates across Asia, the Americas, Europe and Australia, with many of these species having not been characterized. Although these genera are believed to contain well over 400 species of fungi, there are a few standout examples which are revered for their medicinal potential or unusual host pathogenesis.
Ophiocordyceps sinensis, found in the mountains of Tibet, infects and kills ghost moth larvae to give the highly prized herbal remedy "dong chong xia cao," which is believed to treat a plethora of disorders [1]. This prized specimen is identified by the fruiting body growing from the ground, as the infected ghost moth larvae dies situated just below the surface of the soil with its head oriented upward, from which the fruiting body emerges. Ophiocordyceps unilateralis, also known as the zombie-ant fungus, is noted for its pathogenic process in ants, which is characterized by particular behaviour modifications in the host, that leaves the host ant perished with its jaw clamped to a leaf in prime location for spore dispersal [2]. Cordyceps militaris, which is a common component of supplements as it is also believed to have medicinal potential, is often used as a cheaper and more readily available version of Ophiocordyceps sinensis [3,4].
The genome of a handful of these fungi have been sequenced, however, the available assemblies are often fragmented in over 500 contigs [5][6][7][8]. These assemblies do indicate that these fungi are capable to producing natural products, possibly over 30 distinct molecules per species. Only a few of these natural products from entomopathogenic fungi have been isolated and described, including the immunosuppressant, cyclosporine, from Tolypocladium inflatum, and fingolimoid, the immunomodulatory molecule derived from the Isaria sinclarii natural product myriosin, signifying that these fungi may be an underexplored source of novel molecules [9,10].
Natural products have been established as a source of bioactive molecules, however, discovery has dwindled implying the need for new sources. Fungi have been shown to produce a wealth of diverse molecules [11] suggesting that Cordyceps and Ophiocordyceps could be an understudied and relevant avenue for the discovery of natural products. Furthermore, by studying the secondary metabolites in Cordyceps and Ophiocordyceps not only is there the potential for discovery of new bioactives, but for the identification and study of chemicals that have a role in the process of host pathogenesis, from behaviour modifying molecules to insecticides [12]. As genome sequencing becomes cheaper, faster, and more readily available, the possibility of taking a computational approach to genome mining for secondary metabolite discovery in fungi becomes a more realistic possibility, allowing for the study of secondary metabolites which may be cryptic under typical laboratory conditions [13]. Indeed, laboratory culture of these organisms is a challenge due to their slow growth ratesa genome-based approach to their natural product genes is likely essential for this field to progress.
In this study, a new method of long read sequencing, Pacific Biosciences SMRT sequencing is applied to an exemplary sample from the Cordyceps genera, Cordyceps militaris, a strain isolated from butterfly pupa. The overarching goal is to provide a chromosome level genome assembly to serve as a model for the genera. Furthermore, as these fungi have the potential to produce many understudied natural products, this study is focused on the genetic potential for secondary metabolite expression in this organism.

General genome features
Purified genomic DNA isolated from culture of C. militaris grown up from a single colony was sequenced using the Pacific Biosciences platform using a sheared large insert library [14]. Sequence data from 6 SMRT cells, providing approximately 180× coverage were assembled using two de novo assemblers, Celera with the PBcR protocol and the HGAP2 protocol from SMRT portal [15][16][17]. Both chosen assemblers were applied to self-correct the reads, a process in which the shorter PacBio reads were used to error correct the long PacBio reads. These corrected reads were then subsequently assembled into contigs. The PBcR-pipeline gave an assembly with 32 contigs, four of which had telomeric repeats (CCCTAA or TTAGGG) n on either the 5′ or 3′ end of the contig.
The second assembly protocol, the HGAP2 protocol from the SMRT portal software package, which also included an additional polishing step using the Quiver algorithm, yielded an even further improved assembly. This assembly, which contained 18 contigs, gave five assembled chromosomes, having telomeric repeats (CCCTAA or TTAGGG) n on both the 5′ and 3′ ends of the sequence and four having telomeric repeats on one of the 5′ or 3′ end of the sequence. After manually curating the assembly and submitting the curated assembly to the SMRT resequencing protocol, the resultant assembly contained 7 contigs, each with telomeres on both ends, indicative of 7 chromosomes. Coverage across these seven chromosomes, including regions where the assembly was manually curated is shown in Additional file 1: Figure S1. The coverage across the chromosomes is generally consistent in the assembly after manual curation and the SMRT resequencing protocol. The N 50 (5.78kba) and N max (8.29kba) remain unchanged when comparing the initial assembly and the curated and resequenced assembly. However, there is a spike in coverage in the contig corresponding to chromosome IV, possibly implying a collapsed repeat Insect pathogenic fungi are highlighted by a blue box region within the assembly (Additional file 1: Figure S1D). When this area of the assembly is further inspected, it is noted that there are a large number of short low quality repeated reads, which align on top of >30× coverage of high quality reads that span this repetitive region. The repetitive sequence of this region and the overabundance of low quality reads in this region was noted prior to manual curation. The initial assembly consisted of 3 contigs containing these short repetitive reads, each about 30,000 bp in length, which overlapped in this region. With an overall coverage of approximately 150×, a genome size of 33.6 Mba, and a GC content of 50.9%. (Table 1), the BUSCO completeness of this assembly is 98.2% with only 0.4% missing [18]. Furthermore, the assembly is comparable to the previously sequenced C. militaris Cm01 (though that sequence is broken into a very large number of contigs) with a similar genome size of 32.2 Mba and a GC content of 51.4% [6].
The seven assembled C. militaris chromosomes range in size from 1.9 to 8.3 Mba. The sequenced haploid genomes of Aspergillus niger and Neurospora crassa contain eight and seven chromosomes, respectively [19,20]. Furthermore, a karyotype analysis of Tolypocladium inflatum shows that this related species contains seven chromosomes ranging in size from 1.0 to 6.3 Mba [21], suggesting that the assembly with seven chromosomes is reasonable for C. militaris.
The MAKER genome annotation pipeline [22,23] predicted 9907 genes for C. militaris. Passing the MAKER gene predictions through an additional evidence modeler using Funannotate gave a set of 9371 genes with a BUSCO analysis of the resulting gene set estimating a completeness of 93.7% with 3.9% of genes missing [18,24]. Estimates of mean gene length, mean exon length, mean intron length and gene density ( Table 2) are similar to those of C. militaris Cm01 and other filamentous ascomycete fungi (Table 3) [5-7, 12, 25, 26]. An Interpro analysis of the annotated genes using the Blast2GO suite was used to assign 8792 genes (93.8%) of genes InterPro IDs. A Gene Ontology (GO) annotation was assigned to 6453 of the genes (68.9%).

Mating type loci
The sequence of our isolate revealed the presence of only a MAT 1-2-1 mating type gene present on chromosome VII, supporting the notion that this fungus is indeed heterothalic. Both MAT 1-1, MAT 1-2 and strains with hybrid mating loci have been reported [6]. The previously sequenced C. militaris Cm01 strain was determined to have both MAT 1-1-1 and MAT 1-1-2 mating type genes. The C. militaris Cm06 strain was reported to be a hybrid strain containing both the MAT 1-1 and MAT 1-2 mating types, with single spore isolates from this hybrid strain producing progeny with either the MAT 1-1 (93.3%) or MAT 1-2 (6.7%) loci [6]. Both the MAT 1-1 and MAT 1-2 containing isolates have been shown to fruit, but only the hybrid strain containing both the MAT 1-1 and MAT 1-2 loci was able to produce mature spores [6]. However, fruiting bodies were not observed under analogous conditions with the ATCC® 34164 strain. Comparison of the genomic regions containing the mating type genes in both our ATCC® 34164 strain and the Cm01 strain reveal that the genes in these regions are highly similar, with the exception of the MAT genes (Fig. 2).

Cordycepin
One hallmark molecule of interest in C. militaris is the nucleoside analogue cordycepin. Although this biosynthesis is unknown, it is proposed that this mechanism is dependent upon a reduction step, believed to be potentially catalyzed by a ribonucleoside diphosphate reductase (RNR) [6,27,28]. However, our sequenced C. militaris is similar to the sequenced Cm01 strain, in that it only seems to possess two type I RNRs (genes A9K55_000536 and A9K55_003140), both of which have homologues in non-cordycepin producing fungus, and have been identified in C. militaris Cm01, leaving the biosynthesis of cordycepin elusive [6].

Secondary metabolite potential
Sequenced fungi from the Cordyceps, Ophiocordypces, and related genera have revealed the potential for production of over 30 diverse secondary metabolites per strain [7,8,26,29,30]. The limited number of prior systematic studies to identify bioactive secondary metabolites from Cordyceps and related fungi have shown that a number of novel molecules can be produced by these microbes [31][32][33][34][35]. However, these studies do not nearly capture the full secondary metabolite potential of these fungi, likely due to the fact that many of these metabolites may be cryptic and not expressed under the tested laboratory conditions. To determine whether this fungus could produce a wealth of secondary metabolites, the genetic potential for diverse metabolite production became the focus of the study. Using two gene cluster predictors, Anti-SMASH and SMURF, all seven C. militaris chromosomes were profiled for the presence of genes that could be responsible for the biosynthesis of secondary metabolites [36][37][38]. The fungal version of Anti-SMASH, predicted 32 secondary metabolites and SMURF predicted 25. Taken together, the two algorithms predicted the presence of 36 unique gene clusters which could be responsible for secondary metabolite production in C. militaris ( Table 4). Distribution of these secondary metabolite producing genes were mapped on the 7 chromosomes (Fig. 3a). No gene clusters for secondary metabolites were noted in the presumably collapsed area of the genome shown in chromosome IV. The 36 metabolite producing gene clusters were from a variety of classes, including eight nonribosomal peptide synthetases (NRPS), seven type 1 polyketide synthases (T1PKS), six polyketide synthasenonribosomal peptide synthetase (PKS-NRPS) hybrids, four terpenes, one indole and ten falling into other classes (Fig. 3b).
For comparison with the previously sequenced strain, the antiSMASH algorithm was also used to predict the presence of natural product producing    gene clusters in the in C. militaris Cm01. In Cm01 there were 28 natural product clusters identified by AntiSMASH, compared to the 32 in our strain of study. All of the 28 clusters were present in the ATCC® 34164 strain and the ATCC®34164 strain had four additional natural product clusters identified. These additional clusters are predicted to produce one indole (V-3), two T1PKS (VI-2, VI-8) and one T1PKS-NRPS (VII-5). Furthermore, using the Clus-terFinder algorithm in the fungal version Anti-SMASH, an additional 41 putative clusters were predicted, bringing the total to 73 predicted clusters from AntiSMASH with ClusterFinder, suggesting that the secondary metabolite potential of this organism is impressive.

Discussion
Herein is described the first chromosome level assembly of a Cordyceps genome. This seven chromosome assembly has revealed that this heterothallic strain, which contains 9371 genes, is capable of producing a wealth of secondary metabolites. Of the 36 gene clusters identified in ATCC®34164 by both the antiSMASH and SMURF algorithms (Additional file 2: Table S1), 3 clusters, III-1, V-6 and VII-5, are of particular interest as they have homology with gene clusters from other organisms that produce characterized natural products [39,40]. It seems that a logical product of cluster III-1 could be a 2-pyridone alkaloid molecule. The hybrid NRPS-PKS central to this cluster (A9K55_001190) is similar to the NRPS-PKS responsible for the production of desmethylbassianin (70% identity) and tenellin (67% identity) from Beauveria bassiana and fumosorinone (66% identity) from Isaria fumosoinone [41][42][43]. Both the desmethylbassanin and tennellin gene clusters consist of 4 genes: the NRPS-PKS hybrid, an enoyl reductase and two cytochrome P450s (Fig. 4). In C. militaris, based on the sequence of gene A9K55_001191 it seems that the missing enoyl reductase may be fused with the cytochrome P450. Interestingly, a structurally related pigmented derivative, militarinone A, and the variants militarinone B-D, have been isolated from the possible C. militaris anamorph, Paecilomyces militaris [44,45]. However, militarinone A-D was not identified by mass in extracts of the C. militaris strain of interest in this study.
On chromosome V, a cluster (V-6) with homology to the emercellamide producing cluster is present (Fig. 5). The emercellamide family molecules produced from the hybrid NRPS-PKS containing cluster have been described in the marine fungus Emericella, as well as the fungus Aspergillus nidulans [46,47]. Other related molecules, scopularide A and W493-B have been isolated from Scopulariopsis brevicaulis and Fusarium pseudograminearum, respectively [48][49][50]. The biosynthesis of emericellamide A in Aspergillus has been described and is shown to rely on four genes [47]. Comparing this emercellamide-like cluster in C. militaris to the gene cluster producing emercellamide in Aspergillus shows a conservation of 4 genes: an NRPS, a PKS, an acyltransferase and a CoA ligase. The NRPS present in C. militaris (A9K55_005039) has 98% coverage and 43% identity with the NRPS in Aspergillus nidulans.
A cluster with similarities to the equisitin-producing cluster (VII-5) is also present in C. militaris on chromosome VII. This molecule, equisetin, was described as having structural similarities to the cholesterol lowering molecule lovastatin and was first isolated from Fusarium equiseti with a described bioactivity as a HIV-1 integrase inhibitor [51][52][53]. The biosynthesis, studied in Fusarium b a Fig. 3 Putative natural products in C. militaris. The classes of natural products are denoted by the following colors NRPS (cyan), T1PKS (orange), T1PKS-NRPS (red), Indole (indigo), Other (green) Terpene (violet). a The distribution of natural product producing gene clusters on the chromosomes of C. militaris. b The number of natural product gene cluster on each chromosome, grouped by class heterosporum, reveals that the gene cluster consists of seven genes, two NRPS/PKS, two regulators, an oxidase, a methyltransferase and a transporter [53]. The comparable cluster in C. militaris consists of five genes, homologous to the genes present in the F. heterosporum minus one of the regulators and the oxidase (Fig. 6). The NRPS/PKS present in C. militaris (A9K55_008762) has 99% coverage and 50% identity with the NRPS/PKS in Fusarium heterosporum and seems to be well conserved among fungi in the Aspergillus and Penicillum genera.
This assembly of a genome from the Cordyceps genera, Cordyceps militaris, has shown the potential for production of an array of potentially novel natural products. This species is predicted to produce at least 36 secondary metabolites, three of which have significant similarity to characterized gene clusters. As fungal secondary metabolites can be cryptic under standard laboratory conditions, this assembled genome will allow for the application of genome mining techniques to guide the discovery and identification of new natural products. This can progress forward through a variety of techniques; one approach is to heterologously express gene clusters identified in the C. militaris genome. Alternatively, utilizing the genome to extrapolate potential natural products for expression can give important clues about the structure and favorable culture conditions of a secondary metabolite associated with a characterized Fig. 4 Comparison of gene clusters producing a 2-pyridone alkaloid molecule to the genes A9K55_001190 to A9K55_001192 in cluster III-1 in C. militaris ATCC ® 34164. The desmethylbassianin cluster (dmb), tenellin (ten), and fumosorone (fum) clusters are shown. Genes are color coded by function. Structures of 2-pyridone alkaloids are displayed Fig. 5 Comparison of genes A9K55_005039, A9K55_005040, A9K55_005043 and A9K55_005044 in C. militaris ATCC® 34164 to an emericellamide producing gene cluster in Aspergillus nidulans. Genes are color coded by function. The structure of emericellamide is displayed Fig. 6 Comparison of genes A9K55_008762 to A9K55_008769 in C. militaris ATCC® 34164 to an equisetin producing cluster in Fusarium heterosporum. Genes are color coded by function. The structure of equisetin is displayed gene cluster. This knowledge can increase the likelihood of the production of the correlated molecule and simplify structural determination. Regardless, this assembly has shown that there is a great potential for the production of secondary metabolites in C. militaris and that this and other fungi from related Cordyceps and Ophiocordyceps genera could provide a wealth of molecular structural diversity.

Conclusions
Presented here is the first chromosome level assembly of a genome from the Cordyceps genera. This assembly and analysis has revealed that C. militaris has seven chromosomes containing a wealth of gene clusters for secondary metabolite production. Of the 36 gene clusters identified using the antiSMASH and SMURF algorithms, three clusters are found to have a high degree of similarity with clusters from other organisms that produce a known molecule. With this genome, further study and characterization of the secondary metabolites produced by C. militaris can be aided through genome based techniques including heterologous expression of gene clusters. As there is great potential for the production of secondary metabolites from C. militaris, this is one step towards discovering and characterizing the wealth of molecular structural diversity in this genera.

Phylogenetic tree construction
To compare the fungal species of interest, 18S rRNA sequences were obtained from the Silva database [54]. Sequence alignment was performed using ClustalW [55]. To construct the phylogenetic tree, The evolutionary history was inferred using the Neighbor-Joining method [56]. The optimal tree with the sum of branch length = 0.35499733 is shown. The evolutionary distances were computed using the Maximum Composite Likelihood method [57] and are in the units of the number of base substitutions per site. The analysis involved 12 nucleotide sequences. Codon positions included were 1st + 2nd + 3rd + Noncoding. All ambiguous positions were removed for each sequence pair. There were a total of 988 positions in the final dataset. Evolutionary analyses were conducted in MEGA7 [58].

Fungal strain and maintenance
Cordyceps militaris ATCC® 34164 was received from the American Type Culture Collection (ATCC). This strain, as described in the ATCC records was isolated from a butterfly pupa. Fungal cultures were maintained at 23.0°C on potato dextrose agar. The nrDNA of extracted genomic DNA was amplified using the ITS4 and ITS5 primer pairs, sequenced and compared against the BLAST database to determine sample validity [59][60][61].

Genomic DNA extraction and purification
Liquid cultures, containing 5 mL of seed media (10 g peptone, 40 g maltose, 10 g yeast extract, 1 g agar in 1 L DI water) in a culture tube were aseptically inoculated with a 3 mm square agar slab containing mycelial growth. The fungal mycelia were grown for 5 days at 23.0°C. The mycelial mat was harvested, rinsed with sterile TE buffer and then frozen in liquid nitrogen. Frozen mycelia were macerated with a mortar and pestle with a spatula tip of aluminum oxide to aid in grinding the sample. The mycelial powder was transferred to a set of epitubes and 500 μL of CTAB DNA extraction buffer was added (100 mM Tris pH = 8.0, 10 mM EDTA, 2% CTAB, 2.8 M NaCl). The samples were incubated at room temperature for 5 min, then 2 μL of RNAse A (Thermo Scientific, 10 mg/mL) and 10 μL of Proteinase K were added (Invitrogen, 10 mg/mL) and inverted to mix. After centrifuging for 5 min, the pellet was ground in the epitube with a pellet pestle, then incubated for an additional 5 min before purification with phenolchloroform. Each sample was washed twice with phenolchloroform (50:50, phenol buffered with Tris pH 8.0, 600 μL) then twice with chloroform (600 μL). The resulting DNA containing aqueous portions were pooled and DNA was precipitated using cold ethanol (2.5× sample volume) and 3 M sodium acetate (0.1× sample volume). DNA was precipitated for at least 30 min by storing at −20°C. The DNA precipitate was collected by centrifuging for 30 min, the pellet was washed with 70% ethanol and resuspended in TE buffer. The DNA was further purified with AMPure XP beads (Agencourt) by using an equal volume of beads to volume of DNA and eluting into TE. DNA was quantified using a PicoGreen assay (ThermoFisher) prior to sequencing.

Genome sequencing
The Cordyceps militaris DNA was sequenced using Pacific Biosciences RS II sequencing at the Genome Quebec Innovation Center (McGill University, Montreal, Canada). The sample was prepared using a large insert sheared DNA library and was sufficient for sequencing 6 SMRT cells.

Genome assembly
The sequencing reads were assembled using two different assemblers. The first assembler chosen was the PBcR pipeline from the Celera assembler (version 8.3rc2) using a genome size of 32 Mba [16]. The second assembler was SMRT portal (version 2.3.0) launched from an Amazon machine image. Assembly was performed on all 6 SMRT cells using the RS_HGAP Assembly.2 application with default settings and a genome size of 32 Mba [62]. The resulting assembly yielded 18 contigs, with five of these contigs containing characteristic telomeric (CCCTAA) n or (TTAGGG) n repeats on both ends and four of these contigs containing telomeric repeats on one end. The 5 contigs with telomeres on both ends were taken to be fully sequenced chromosomes. The remaining 13 contigs were evaluated for overlapping regions that could possibly be used to join the contigs. Two of these 13 contigs were discarded due to low coverage (<30×), three of these remaining contigs were found to have overlapping regions that allowed them to be joined into 1 supercontig and the remaining 8 contigs were found to also contain overlapping regions that would allow them to be joined into a second supercontig. These overlapping regions were evaluated by subjecting the entire genome of 5 chromosomes initially assembled by SMRT portal, plus the two manually curated supercontig chromosomes, to the SMRT portal resequencing protocol as a reference genome, along with manually evaluating these overlapping regions by evaluating the reads that spanned the overlapping regions. The resulting assembly was in 7 contigs, with each end of the contig terminating in a telomeric repeat sequence.

Gene prediction, functional annotation and protein classification
Genome annotation was performed using the MAKER (version 2.31.8) pipeline using three ab inito gene prediction methods: Augustus trained for Fusarium graminearum, and GeneMark-ES and SNAP selftrained on the C. militaris genome [22,23,[63][64][65]. Protein data from Cordyceps brongniartii, C. militaris, Cordyceps confragosa, Ophiocordyceps sinensis, Ophiocordyceps unilateralis and Tolypocladium ophioglossoides were used as protein evidence in MAKER. EST from Cordyceps militaris were downloaded from Gen-eBank and used as EST evidence in MAKER. Repeat elements were identified using Repeat Masker using the Repbase Library 20150807 [66]. A final set of consensus gene predictions was chosen using Exonerate [67]. The final gene set from MAKER was subjected to additional evidence modeling using Funannotate (0.7.0) [24]. The gene models were functionally annotated using the BLAST component of the Blast2GO software package and searching against the NCBI nr protein database (accessed July 2017) with the best hit being selected [68]. Gene families were established using the Interpro database using Blas-tProDOM, HMMPIR, HMMPfam, SuperFamily, Sig-nalPHMM, HMMPanther [69]. The BLAST hits were mapped to the Gene Ontology database and KEGG analysis was carried out [70]. Secondary metabolite genes and gene clusters were predicted using both AntiSMASH, fungal version 4.0.0 and SMURF (accessed September 2016) [36,38].

Attempt at fruiting body production
Using the protocol outlined in Zheng et al. fruiting body production was attempted with C. militaris ATCC® 34164 [6].

Attempt at militarinone production
Using the protocol in Schmidt et al. milititarone A-D production was attempted [44]. Specifically, precultures of C. militaris were used to inoculate 150 mL of medium (2% glucose, 2% neopeptone, 0.5% glycine, 0.2% K 2 HPO 4 , MgPO 4 -7H 2 O) in still 500 mL Erlenmeyer flasks at 25°C. After 20 days the broth was removed and the mycelia were collected and freeze dried, then extracted with methanol. This methanol extract was treated with water (1.5 mL for 10 g of extract) and then partitioned in a 1:1:1:1 mixture of ethyl acetate/methanol/hexane/1% acetic acid. The lower phase was collected, concentrated, and then analyzed for militarinone production via UPLC-MS. No mass peaks corresponding to ionized militarinone A-D or sodium or acetate adducts of those natural products were apparent.

Coverage and identity
Coverage and identity of C. militaris genes compared to genes in known biosynthetic clusters was determined using BLAST [61].

Additional files
Additional file 1: Figure S1. Coverage across chromosomes. Coverage across chromosomes from SMRT analysis resequencing protocol assembly. (DOCX 1935 kb) Additional file 2: Table S1. Predicted gene clusters in C. militaris. Predicted gene clusters are labeled, putative natural product class and the predicted length of each enzyme that is part of the putative cluster is given. (DOCX 106 kb) Abbreviations HGAP: Hierarchical genome assembly process; MAT: Mating type; PacBio: Pacific Biosciences; PBcR: PacBio corrected reads