Analysis of the Indole Diterpene Gene Cluster for Biosynthesis of the Epoxy-Janthitrems in Epichloë Endophytes

Epoxy-janthitrems are a class of indole diterpenes with structural similarity to lolitrem B. Two taxa of asexual Epichloë endophytes have been reported to produce epoxy-janthitrems, LpTG-3 (Lolium perenne Taxonomic Group 3; e.g., NEA12) and LpTG-4 (e.g., E1). Epichloë epoxy-janthitrems are not well understood, the biosynthetic pathway and associated gene complement have not been described and while the literature suggests they are associated with superior protection against pasture insect pests and are tremorgenic in grazing mammals, these properties have not been confirmed using isolated and purified compounds. Whole genome sequence analysis was used to identify candidate genes for epoxy-janthitrem biosynthesis that are unique to epoxy-janthitrem producing strains of Epichloë. A gene, jtmD, was identified with homology to aromatic prenyl transferases involved in synthesis of indole diterpenes. The location of the epoxy-janthitrem biosynthesis gene cluster (JTM locus) was determined in the assembled nuclear genomes of NEA12 and E1. The JTM locus contains cluster 1 and cluster 2 of the lolitrem B biosynthesis gene cluster (LTM locus), as well as four genes jtmD, jtmO, jtm01, and jtm02 that are unique to Epichloë spp. that produce epoxy-janthitrems. Expression of each of the genes identified was confirmed using transcriptome analysis of perennial ryegrass-NEA12 and perennial ryegrass-E1 symbiota. Sequence analysis confirmed the genes are functionally similar to those involved in biosynthesis of related indole diterpene compounds. RNAi silencing of jtmD and in planta assessment in host-endophyte associations confirms the role of jtmD in epoxy-janthitrem production. Using LCMS/MS technologies, a biosynthetic pathway for the production of epoxy-janthitrems I–IV in Epichloë endophytes is proposed.


Introduction
Perennial ryegrass (Lolium perenne L.) is often infected with fungal endophytes that include sexual Epichloë species and their asexual derivatives. In the case of asexual Epichloë spp. endophytes, the host grass-endophyte interaction is considered mutualistic as both the plant and the fungus receive benefits [1]. The plant provides the fungus with a means for survival and reproduction and the fungus protects this environment by producing compounds that aid the defence of the host plant. While the Lolium-Epichloë interaction produces compounds beneficial for plant growth and persistence, in a Janthitrems were first isolated from Penicillium janthinellum strains obtained from pastures in which ryegrass staggers had been observed [19,20]. Penicillium species, which produce a number of tremorgenic mycotoxins, were thought to be the cause of ryegrass staggers until it was later proven to be Epichloë-derived lolitrem B [19,21]. The first evidence that janthitrems and lolitrem B share a common pathway came from an experiment in which P. janthinellum provided with C 14-labelled paxilline showed incorporation of the radiolabel into janthitrem B, indicating that paxilline is a likely precursor for janthitrem B biosynthesis [22,23]. Schardl et al. [24] suggested that, for Epichloë endophytes, janthitrems are synthesised from terpendole I, which is also a precursor of lolitrem B biosynthesis. Tapper et al. [11] first identified the novel janthitrem named 11,12-epoxy-janthitrem G (epoxy-janthitrem I) from AR37 endophyte-infected perennial ryegrass. Epoxy-janthitrem I is the major janthitrem alkaloid produced by perennial ryegrass endophytes.
Indole diterpenes are a diverse group of secondary metabolites produced by filamentous fungi. The best characterised are from Penicillium, Aspergillus, Claviceps and Epichloë species; however, the number of species in which indole diterpenes have been identified continues to increase [25]. The indole diterpenoid compounds comprise a cyclic diterpene-derived skeleton (derived from four isoprene units, i.e., geranylgeranyl diphosphate (GGDP)) and an indole moiety (derived from tryptophan or a tryptophan precursor) [26,27]. The structural diversity of these metabolites is due to Janthitrems were first isolated from Penicillium janthinellum strains obtained from pastures in which ryegrass staggers had been observed [19,20]. Penicillium species, which produce a number of tremorgenic mycotoxins, were thought to be the cause of ryegrass staggers until it was later proven to be Epichloë-derived lolitrem B [19,21]. The first evidence that janthitrems and lolitrem B share a common pathway came from an experiment in which P. janthinellum provided with C 14labelled paxilline showed incorporation of the radiolabel into janthitrem B, indicating that paxilline is a likely precursor for janthitrem B biosynthesis [22,23]. Schardl et al. [24] suggested that, for Epichloë endophytes, janthitrems are synthesised from terpendole I, which is also a precursor of lolitrem B biosynthesis. Tapper et al. [11] first identified the novel janthitrem named 11,12-epoxy-janthitrem G (epoxy-janthitrem I) from AR37 endophyte-infected perennial ryegrass. Epoxy-janthitrem I is the major janthitrem alkaloid produced by perennial ryegrass endophytes.
Indole diterpenes are a diverse group of secondary metabolites produced by filamentous fungi. The best characterised are from Penicillium, Aspergillus, Claviceps and Epichloë species; however, the number of species in which indole diterpenes have been identified continues to increase [25]. The indole diterpenoid compounds comprise a cyclic diterpene-derived skeleton (derived from four isoprene units, i.e., geranylgeranyl diphosphate (GGDP)) and an indole moiety (derived from tryptophan or a tryptophan precursor) [26,27]. The structural diversity of these metabolites is due to the different patterns of ring substitutions, different ring stereochemistry and additional prenylations [27]. In the context of their natural ecosystem, indole diterpenes act as bioprotectants, defending the fungus or protecting the host plant against herbivory by animals and insects. Many are potent mammalian toxins, and some have been shown to have antimicrobial and insecticidal properties [28]. Thus, production of indole diterpenes offers evolutionary advantages for the producing organism and its host plant.
While some literature on the properties of epoxy-janthitrem-producing endophytes exists, janthitrems themselves are not well understood when compared to other alkaloid groups synthesised by Epichloë endophytes. For example, while the genes and biosynthetic pathway for lolitrem B [34,35,38] and ergovaline [39] have been described and extensively characterised, the epoxy-janthitrem biosynthetic pathway and associated gene complement have not been described.
The literature suggests epoxy-janthitrems are associated with superior protection against pasture insect pests and are tremorgenic to grazing mammals, however, these properties have not been confirmed using isolated and purified compounds. This is in part because they are too unstable to isolate and purify in large enough quantities to study [15]. If the genetic basis for epoxy-janthitrem production is understood, it then becomes possible to characterise their properties. In this study, we identify the JTM locus (epoxy-janthitrem biosynthetic gene cluster) and the genes required for epoxy-janthitrem biosynthesis in asexual Epichloë endophytes. We functionally characterise the gene jtmD, using RNAi-mediated gene knockdown followed by liquid chromatography-mass spectrometry (LC-MS) analysis of transgenic endophytes in planta, to confirm the role of jtmD in epoxy-janthitrem I production. Finally, we propose a parsimonious pathway for biosynthesis of epoxy-janthitrems by Epichloë endophytes.

Fungal Strains and Growth Conditions
The properties of Epichloë endophytes used in this study are shown in Table 1. Endophyte cultures were grown either on potato dextrose agar (PDA) (Sigma-Aldrich, St. Louis, MO, USA), or in potato dextrose broth (PDB) at 22 • C, 150 rpm in the dark for a period of 7-10 days depending on growth rate. Genomic DNA was extracted from lyophilized endophyte mycelia using a cetyltrimethylammonium bromide (CTAB) based extraction method [40], and the quality and quantity of the DNA was assessed by agarose gel electrophoresis and specific absorbance measurements using NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, Walthman, MA, USA). Genomic DNA was fragmented in a Covaris instrument (Woburn, MA, USA) to an average size of 100-900 bp. For each endophyte DNA sample, paired-end libraries with inserts c. 400 bp in size were prepared using the standard protocol (TruSeq DNA Sample Prep V2 Low Throughput; Illumina Inc., San Diego, CA, USA) with paired-end adaptors. Library quantification was performed using the KAPA library quantification kit (KAPA Biosystems, Wilmington, MA, USA) and sequenced using the HiSeq2000 platform (Illumina Inc., San Diego, CA, USA) following the standard manufacturer's protocol.

DNA Extraction and Long-Read Sequencing on PacBio and MinIon Sequencing Platforms
Genomic DNA was extracted from endophyte mycelia snap frozen in liquid N 2 using a modified CTAB extraction method [41]. The quality and specific absorbance measurements of DNA was assessed as explained above. DNA Quantification was measured by Qubit (Thermo Fisher Scientific, Walthman, MA, USA) according to manufacturer's instructions. Genomic libraries were prepared and sequenced according to 20 kb SMRTbell library preparation kit (Pacific Biosciences, Menlo Park, CA, USA) and ligation sequence preparation kit (SQK-LSK109; Oxford Nanopore, Oxford, UK) for PacBio and MinIon, respectively.

Sequence Assembly and JTM Locus Identification
PacBio generated long-reads of NEA12 were filtered, assembled (assuming the genome size as 35 Mb), polished using SMRT Portal software HGAP assembler [42], and split into 500,000 bp pieces for read mapping. For MinIon generated reads of E1, sequence correction, trimming and assembly was performed using the long-read assembler Canu v.1.8 [43], assuming the genome size of 35 Mb. Scaffolded assembly produced by Canu was polished with genome assembly improvement and variant detection tool Pilon v. 1.23 [44], using short-reads generated from Illumina HiSeq 2000 sequencing platform. Final E1 genome assembly was split into 500,000 bp pieces for read mapping. Illumina generated sequence reads were quality trimmed using the Gydle nuclear program (Gydle Inc., Québec City, QC, Canada) such that all reads ≥50 bases long were retained using a base quality cut-off value of 20. All high-quality reads for a given endophyte strain were mapped to the NEA12 PacBio contig 3 (247,475 bp) using the Gydle 'nuclear' aligner version 3.2.1 (Gydle Inc., Québec City, QC, Canada). Reads were mapped with settings: l 50 (length of overlap); s 25 (sensitivity); k 13 (kmer length); m 6 (maximum number of mis-matches); F 3 (filter settings). Alignments were visualised with Gydle program vision version 2.6.14 (Gydle Inc., Québec City, QC, Canada).
The JTM locus was identified, first by identifying the location of the jtmD gene and then annotating the region using a combination of both Augustus [45], gene prediction and manual annotation using the known gene sequences of LTM genes [35,38], and jtmD. Each of the predicted genes were then characterised by BLASTp analysis and comparisons with genes from other indole diterpene gene clusters performed by multiple sequence alignment using MUSCLE v. 3.8.31 [46], with default parameters. To reconstruct tree topology ML analysis was used as implemented in MEGA X [47], using the maximum likelihood method and JTT matrix-based model [48], using the partial deletion option (all positions with less than 95% site coverage were eliminated). The tree with the highest log likelihood is shown. Nearest-neighbour-interchange (NNI) was performed as a heuristic method. The internal branch support was assessed by a search of 500 bootstrapped sets of data.

Transcriptome Expression of Genes Located Within the JTM Locus
Endophyte devoid (E − ) perennial ryegrass cultivar Alto seeds, harvested from endophyte devoid parent plants, and those containing the Epichloë endophyte strains SE and NEA12 were obtained from New Zealand Agriseeds (NZA), Christchurch, New Zealand. Seeds were prepared, RNA extracted and RNA libraries constructed as described [49]. Four biological replicates were prepared for each endophyte strain. The Impact 04 -E1 transcriptome is described by Shinozuka et al. [50]. Raw sequence reads were trimmed using the Gydle 'nuclear' program and all high-quality reads were mapped to E1 JTM locus as detailed above (Gydle Inc., Québec City, QC Canada).

Fungal and Bacterial Strains and Culture Conditions
Epichloë endophyte strain E1, was the endophyte strain of choice due to the toxin profile (Table 1) and fast in vitro growth rate, for jtmD knockdown. Escherichia coli strain, DH5α, was used for construction and maintenance of plasmid constructs (Thermo Fisher Scientific, Walthman, MA, USA) and were grown at 37 • C on Luria-Bertani (LB) agar (5 g yeast extract, 10 g tryptone, 10 g NaCl in 1 L of ddH 2 O) plates or broth. PDA and LB media were supplemented with appropriate antibiotics (50 µg kanamycin, 50 µg/mL spectinomycin).

Plasmid Construction
The plasmid, pDONR 221 and gene cassettes were supplied/synthesised by Invitrogen (Thermo Fisher Scientific, Walthman, MA, USA). Gene cassettes contained inverted repeats of jtmD from E1 (jtmD_129 bp and jtmD_432 bp) separated by a 147 bp spacer (cutinase gene intron from Magnaporthe grisea), flanked by attB1 and attB2 recombination sites. The Gateway™-enabled destination vector (pEND0002) was supplied by Agriculture Victoria Research (Thermo Fisher Scientific, Walthman, MA, USA) [49]. To generate entry clones, gene cassettes (inverted repeats of candidate gene sequences, jtmD_129 bp and jtmD_432 bp, separated by the 147 bp spacer and containing attB1 and attB2 sites), were cloned into the pDONR 221 vector using BP clonase (Thermo Fisher Scientific, Waltham, MA, USA) according to manufacturer's instructions. The final RNA silencing vectors were produced by LR clonase reaction, following manufacturer's recommendations (Thermo Fisher Scientific, Walthman, MA, USA), between the entry vectors carrying the jtmD candidate gene sequences (jtmD_129 bp and jtmD_432 bp) and the Gateway™-enabled destination vector, pEND002. The structures of all expression clones were verified by restriction enzyme analysis and Sanger sequencing.

Isolation of Fungal Protoplasts
Mycelia were harvested, under sterile conditions, by filtration through layers of miracloth lining a funnel and washed 3 times with 30 mL of sterile ddH 2 O. Mycelia were washed with 10 mL of OM buffer (1.2 M MgSO 4 .7H 2 O, 10 mM Na 2 HPO 4 , 100 mM NaH 2 PO 4 .2H 2 O, pH 5.8) and transferred to a sterile 250 mL plastic vessel. Freshly prepared 10 mg/mL Glucanex (30 mL) (Sigma-Aldrich, St. Louis, MO, USA) in OM was added and incubated for 18 h at 30 • C with gentle shaking (80 rpm). The glucanex/protoplast solution (30 µL) was examined under a microscope to confirm successful digestion. Protoplasts were filtered through miracloth in a funnel, into 15 mL sterile glass centrifuge tubes (Gentaur, Belgium), and placed on ice. Each tube was carefully overlayed with 2 mL of ST buffer (0.6 M sorbitol, 100 mM Tris-HCl, pH 8.0) and centrifuged (Avanti ® J-25I; Beckman Coulter, Brea, CA, USA) (5000 rpm for 5 min at 4 • C). Following centrifugation, protoplasts formed a white layer between the glucanex solution and ST buffer, and this layer was carefully removed. STC buffer (1 M sorbitol, 50 mM CaCl 2 ·2H 2 O, 50 mM Tris-HCl, pH 8.0) (5 mL) was added to the protoplast solution in fresh sterile glass tubes. Samples were gently inverted once and centrifuged (5000 rpm for 5 min at 4 • C). Protoplast pellets were pooled with 5 mL of STC buffer and centrifugation was repeated (5000 rpm for 5 min at 4 • C) until only one pellet remained. Excess STC buffer was removed, and the final protoplast pellet was re-suspended in 500 µL of STC buffer. Protoplast concentration was estimated by diluting protoplasts (1/100 and/or 1/1000 in STC buffer) and counting using 40× magnification (Olympus, BX41) and Reichert Bright-Line hemocytometer with improved Neubauer rulings (Hausser Scientific, Horsham, PA, USA). Protoplasts were diluted with STC to 1.25 × 10 8 protoplasts/mL.

LC-MS Analysis
Extracts were analysed using a 100 mm × 2.1 mm Hypersil Gold 1.9 µm HPLC column fitted to a Scientific Vanquish liquid chromatograph (Thermo Fisher Scientific, Walthman, MA, USA). The mobile phase was: A (0.1% formic acid in water) (Thermo Fisher Scientific, Walthman, MA, USA), and B (0.1% formic acid in acetonitrile) (Thermo Fisher Scientific, Walthman, MA, USA), at 0.3 mL/min. Initial conditions were 98% A, before initiating a linear gradient to 0% A over 11 min, and this was maintained for 4 min before returning to the initial gradient conditions. Extracts (3 µL) were injected onto the LC-MS system and analysed using a QExactive Plus mass spectrometer (Thermo Fisher Scientific, Walthman, MA, USA) in positive electrospray ionisation (ESI+) mode over a mass range of 80-1200 m/z. The MS parameters were as follows: Resolution, 35,000; normalised collision energy, 30 V and the maximum ion time, 200 ms. The source heater temperature was maintained at 310 • C and the heated capillary was maintained at 320 • C. The sheath, auxiliary and sweep gases (N 2 ) were 28, 15 and four units, respectively. Spray voltage was set at 3.6 kV. Data-dependent MS/MS (MS 2 ) spectra were acquired in a separate analysis on a selected Alto-E1 control sample, with compounds of interested targeted. The resolution was set at 17,500 with a normalised collision energy of 30 V, maximum ion time of 50 µs and an isolation window of 1 m/z. All LC-MS data were acquired and inspected in Xcalibur Qual Browser v.3.0.63 (Thermo Fisher Scientific, Walthman, MA, USA). The metabolites were defined as MS1 or MS2 level metabolites, according to [51]; the elution time for paxilline was compared with a commercially available standard (HPLC grade ≥98% pure) (Sigma-Aldrich, St. Louis, MO, USA), whilst epoxy-janthitrems I-IV were determined by accurate mass (∆ 0.3-5 ppm) and LC-MS/MS fragmentation patterns matched to the literature [11]. Metabolite abundance (expressed as peak area) was determined for paxilline and epoxy-janthitrems I-IV using Xcalibur LCquan v. 2.7.020 (Thermo Fisher Scientific, Walthman, MA, USA). Statistical differences between the control plants (Alto-E1) and the jtmD hairpins were determined at p < 0.01 (Student's t-test, unpaired, two tailed distribution).

Results and Discussion
Since the identification of the biosynthetic pathway for paxilline in P. paxilli [29][30][31]52], gene functionality in other indole diterpene biosynthetic pathways, including lolitrem B in Epichloë spp. [38], shearinine K and penitrem in Penicillium spp. [33,53], aflatrem in A. flavus [32] and more recently nodulisporic acid in Hypoxylon pulicicidum [54] have been elucidated. These pathways share features in common; orthologous genes that encode enzymes for indole diterpene biosynthesis and genes located in clusters in fungal genomes. More recently, with the availability of whole genome sequence for an increasing number of fungal species, secondary metabolite gene clusters have been observed to be enriched in telomere-proximal regions [55]. In this study, we used this knowledge to identify the JTM locus containing the gene set required for epoxy-janthitrem biosynthesis in Epichloë.
3.1. Identification of a Candidate Gene, jtmD, for Epoxy-Janthitrem Biosynthesis Qualitative variation for the capacity to produce a specific metabolite appears to be largely attributable to the endophyte genotype, and presumably reflects the presence or absence of the relevant biosynthetic genes. Indole diterpenes, epoxy-janthitrems and lolitrem B share common biosynthetic precursors. Genes from the lolitrem B pathway were used to identify candidates for epoxy-janthitrem biosynthesis.
LpTG-3 and LpTG-4 endophytes do not produce lolitrem B in association with perennial ryegrass. However, these taxa contain all the ltm genes except LTM3 cluster genes, ltmE and ltmJ [7,10,56]. Young et al. [57], also observed that LpTG-3 strain AR37 was missing the genes, ltmE and ltmJ. Genome sequence analysis was used to identify candidate genes for epoxy-janthitrem biosynthesis in the NEA12 genome. The protein sequences of LtmE and LtmJ from Standard Endophyte (SE), a LpTG-1 strain with similarity to Lp19 [35], were used as query sequences to search the predicted protein database derived from the NEA12 genome. Using this approach, BLASTp searches yielded putative LtmE and LtmJ protein homologues in the library of predicted NEA12 proteins. A search for candidates for epoxy-janthitrem I production unique to LpTG-3 and LpTG-4 genomes yielded a single LtmE homologue and, therefore, the best likely candidate for further investigation.

Identification of the JTM Locus in the LpTG-3 and LpTG-4 Genomes
The NEA12 and E1 genomes were sequenced using long read sequence technology and the contig containing the putative epoxy-janthitrem biosynthetic gene cluster (JTM locus) was identified using the jtmD gene sequence as a query. The gene content of the NEA12 and E1 contigs containing jtmD was then annotated using a combination of both Augustus gene prediction and manual annotation using the known gene sequences of ltm genes and jtmD [35,38,45] (Table 2, Figure S1).
The JTM locus is located telomere proximal (6,230,165-6,396,207 bp) in the 6,405,645 bp contig 3 of E1. The homologous chromosome in the Epichloë (Fl1) reference genome is the 6 Mb chromosome III [55,58], which harbors the LTM locus. As previously reported for the LTM locus, the JTM locus is characterized by retrotransposon-rich and AT-rich regions [35,38]. The topology of the partial LTM locus in LpTG-3 and LpTG-4 exhibits more similarity to the E. festucae LTM locus than the LpTG-1 LTM locus which has two retrotransposon relics inserted between ltmK and the pks pseudogene. Given the similarity between the LpTG-3 and LpTG-4 loci, it is likely that in addition to these two asexual taxa there is, or once was, a common ancestral sexual Epichloë species that synthesized epoxy-janthitrems.
The pks pseudogene defines the boundary between sequence in common to LpTG-1, LpTG-2, LpTG-3 and LpTG-4 genomes and a previously undescribed genome sequence unique to epoxy-janthitrem producing strains (Figure 3a). This region is characterized by four genes, a transposase with a MULE domain, a Helitron helicase-like transposable element, and AT-rich regions. Two gene clusters, termed Cluster 3 and Cluster 4, were identified; none of the genes in this region have been previously described in Epichloë endophytes.
Cluster 3 contains two genes, jtmD and jtmO. As described above, JtmD exhibits highest homology to aromatic prenyl transferases (Figure 2). JtmO exhibits highest homology to NodO from H. pulicicidum (69.8%) (Table 2; Figure S2A). JtmO also has homology to JanO, a FAD-binding oxidoreductase associated with synthesis of shearinines in P. janthinellum (52%) [33]. Genes with similar predicted functions have been identified in other indole diterpene gene clusters. For example, the paxilline biosynthetic gene cluster of P. paxilli (paxO, previously referred to as PP121) [29] and the penitrem biosynthetic gene cluster of P. crustosum (penO) [33]. Similarly, to these proteins, the JtmO protein product is likely to have a role in modification of the indole diterpene core.
Homologues of JtmD and JtmO in Penicillium species are often found located side by side. It is interesting to note that in the E. weberi genome (GenBank: LGSR01000002.1) [60], the two gene homologues identified in this study (JtmD, predicted protein KOS22745.1; JtmO, predicted protein KOS22754.1) are also found adjacent to each other. Despite having a reduced genome size and gene content in comparison to less specialized relatives, E. weberi retains genes necessary for production of toxins, including indole diterpenes such as shearinines [61,62].
The high identity of JtmD to NodD1, and JtmO to NodO [54], is interesting. Nodulisporic acids are a group of emindole SB-like indole diterpenes that are only produced by a monophyletic lineage of asexual endophytic fungal strains widely distributed in the tropics, H. pulicicidum [54,63]. Nodulisporic acids are of significance because of their highly potent insecticidal activity [63,64]. Cluster 4 contains two predicted genes, jtm01 and jtm02. The Jtm01 protein shows homology to predicted cytochrome P450 proteins in fungi (Table 2; Figure S2B). Jtm01 does not have an ortholog in any other indole diterpene gene cluster characterized to date. There are, however, predicted cytochrome P450 monooxygenases within other indole diterpene gene clusters that appear to be unique to each cluster, for example PJ-13 within the P. janthinellum JAN locus [33], and AF115 in the ATM1 locus [32]. The role of Jtm01 is not clear based on sequence homology, however, as the predicted protein shows some similarity to cytochrome P450 6A1 proteins in other fungi (e.g., Ophiocordyceps sinensis CO18, accession EQL02233; Figure S2B), it may have a role in epoxidation [65].
The Jtm02 protein does not exhibit significant homology to any protein sequences in the GenBank database, however, it does show some similarity to proteins predicted to be membrane bound O-acyl transferase (MBOAT) proteins in other fungi ( Figure S2C). While the role of Jtm02 is not clear based on sequence homology, as an acyl transferase Jtm02 is most likely associated with the acetylation of the janthitrems and epoxy-janthitrems.
Thus, the predicted genes for epoxy-janthitrem biosynthesis are located at a single locus, JTM, within the LpTG-3 and LpTG-4 genomes, as is the case for previously characterized indole diterpenes including lolitrem B [35] and paxilline [31].

Transcriptome Expression of Genes Located in the Janthitrem Biosynthesis Gene Cluster
Following identification of four gene clusters using whole genome sequence analysis, transcriptome analysis of candidate genes for epoxy-janthitrem biosynthesis was performed to study their expression in planta. This was achieved by mapping RNAseq reads generated from perennial ryegrass-endophyte associations, NEA12, E1 and SE to the E1 JTM locus (Figure 3b). Mapping of reads to previously defined Cluster 1 and Cluster 2 genes was observed for all three ryegrass-endophyte associations. In addition, reads were mapped to the genes proposed to be involved in epoxy-janthitrem biosynthesis, jtmD, jtmO, jtm01, and jtm02 only in NEA12 and E1 associations, confirming expression in planta. RNAseq reads generated from the perennial ryegrass-SE association were not mapped to Cluster 3 and Cluster 4 genes further confirming the absence of these two clusters.
Production of the precursory metabolite paxilline was also measured. Paxilline abundance was perturbed in response to jtmD knockdown, however, a significant change in paxilline abundance was only observed when moderate (49%, 64%) reduction of epoxy-janthitrem was observed. When compared to Alto-E1 symbiota, abundance of paxilline was significantly reduced in Alto-E1 jtmD 129bp (p < 0.0001; 55% reduction) and Alto-E1 jtmD 432bp-2 (p < 0.001; 40% reduction) symbiota. Interestingly, when compared to the control symbiota, Alto-E1 jtmD 432bp-1 symbiota produced the lowest levels of epoxy-janthitrem I-IV (p < 0.0001; 82% reduction), yet no change in paxilline abundance was observed in these plants. Effects on upstream components of indole diterpene pathways as result of gene knockdown have not frequently been reported. One such study describes that the disruption of the terP cluster involved in terpendole E biosynthesis resulted in overproduction of upstream alkaloids [36]. The terP gene is involved in the metabolic conversion of terpendole E to its downstream counterparts; thus, terP knockout strains of Agrobacterium tumefaciens resulted in terpendole E accumulation [36].   In planta assessment of paxilline and epoxy-janthitrem I-IV production in Alto-E1 host-endophyte associations with jtmD hairpins. The response denotes arbitrary units, measured as chromatography peak area (peak height by width). Bars represent standard error of the mean (SEM). Significance is measured by Student's unpaired t-test and directly compares the Alto-E1 jtmD RNAi hairpins to Alto-E1. * p < 0.01; ** p < 0.001, *** p < 0.0001. n = the number of individual perennial ryegrass-endophyte associations measured.

Proposed Pathway for Epoxy-Janthitrem Biosynthesis
Indole diterpene gene clusters identified to date are characterized by a core set of four genes for the synthesis of paspaline, and a suite of additional genes that encode multi-functional cytochrome P450 monooxygenases, FAD dependent monooxygenases and prenyl transferases that catalyse various regioand stereo-specific oxidations on the molecular skeleton to generate a diversity of indole diterpene products [33].
A framework for the biosynthesis of the epoxy-janthitrems in Epichloë is described based on analogy with other indole diterpene pathways, identification of biosynthesis intermediates using LC-MS, and the functional analysis of jtmD, which implicated the gene in epoxy-janthitrem biosynthesis ( Figure 5).
Robust high-resolution LC-MS was used to target key metabolites proposed to be associated with the biosynthesis of indole diterpene alkaloids, specifically epoxy-janthitrems (I-IV). The extracted ion chromatograms of the compounds, observed in planta in Alto-E1 symbiota, are illustrated in ( Figure S3). The observed retention times were identified by accurate mass (∆ 0.3-5 ppm) and fragmentation patterns (Table 3).
Saikia et al. [30] showed that four genes, paxG, paxM, paxB and paxC, are required for the biosynthesis of the first stable indole diterpene intermediate, paspaline. Genetic analysis of P. paxilli has established that paxilline biosynthesis then requires two cytochrome p450 monooxygenases, paxP and paxQ which utilise paspaline and 13-desoxypaxilline as their respective substrates [30]. Orthologues of these six genes are found in all paspaline-derived indole diterpene gene cluster loci characterized to date. In P. janczewskii and P. janthinellum, incorporation of labelled β-paxitriol, and not α-paxitriol, into penitrem A and E and janthitrem B and C, respectively, suggests that β-paxitriol is an immediate precursor for the complex indole diterpenes with B-stereochemistry [23].
Genetic evidence suggests that the biosynthesis of epoxy-janthitrems proceeds along the same pathway as lolitrem B in other Epichloë species, involving ltmG, ltmC, ltmM and ltmB in the synthesis of paspaline. Here, we propose that initial steps in the epoxy-janthitrem pathway require LtmP and LtmQ for the demethylation and hydroxylation of paspaline for the synthesis of β-paxitriol. JtmD and JtmO are required for the diprenylation and oxidative cyclisation to construct the distinct A/B ring in the janthitremane indole diterpenes. LtmF and LtmK are required for the prenylation and cyclisation of the epoxy-janthitrems II to IV consistent with the lolitrem biosynthesis ( Figure 5). Jtm01 (P450 monoxygenase) is required for epoxidation in the pathway, which is consistent with previous reports [65]. Jtm02 (membrane bound O-acyl transferase), is most likely associated with the acetylation of the janthitrems and epoxy-janthitrems.  Proposed pathway for epoxy-janthitrem I-IV biosynthesis. The suggested scheme follows the KEGG indole diterpene biosynthetic pathway, illustrating a parsimonious route to epoxy-janthitrem I (11, 12-epoxy-janthitrem G) and its variants (epoxy-janthitrems II-IV). All epoxy-janthitrems were tentatively identified by LCMS/MS and fragment ions were matched [11,67]. Proposed pathway for epoxy-janthitrem I-IV biosynthesis. The suggested scheme follows the KEGG indole diterpene biosynthetic pathway, illustrating a parsimonious route to epoxy-janthitrem I (11, 12-epoxy-janthitrem G) and its variants (epoxy-janthitrems II-IV). All epoxy-janthitrems were tentatively identified by LCMS/MS and fragment ions were matched [11,67]. Table 3. Proposed metabolites involved in the biosynthesis of the paxilline and epoxy-janthitrem I-IV, assessed in planta using Alto-E1 samples by LC-MS/MS. The accurate masses (m/z), retention times (RT) and fragmentation data (product ions, MS 2 ) were acquired in positive ionization mode (M+H). Accurate mass and MS 2 results were compared with theoretical masses and fell within the range of 0.3-5 ppm difference (∆ ppm).

Concluding Remarks
The work described here provides a genetic basis for epoxy-janthitrem biosynthesis in Epichloë endophytes. Genes for epoxy-janthitrem biosynthesis are found clustered at a single locus, JTM, on chromosome III of the genome. Functional analysis of the jtmD gene using RNAi knockdown, revealed reduced abundance of epoxy-janthitrems I-IV, implicating JtmD in epoxy-janthitrem biosynthesis. While the role of JtmD and JtmO can be predicted based on comparison to previously well characterized indole diterpene pathways, further work is required to determine a role of Jtm01 and Jtm02 in epoxy-janthitrem biosynthesis. A biosynthetic pathway based on predicted protein function and comparison with other indole diterpene pathways is proposed. As the genes for epoxy-janthitrem production are now determined, it becomes possible to predict janthitrem production by screening strains for the presence of jtm genes and manipulate the pathway, using genome editing technology. A limitation to the study of epoxy-janthitrems in the past has been their unstable nature once isolated; it is now possible to synthesize sufficient quantities for isolation and use in accurate measurement as well as in insect/animal toxicity studies. There is also potential for creating novel indole diterpenes in Epichloë endophytes using the existing LTM gene clusters and adding new designer gene clusters using a targeted genome editing approach. Deeper understanding of epoxy-janthitrem production enables exploiting their potential use in pasture pest management, enhancing pasture persistence while minimizing the threat posed by their likely tremorgenicity.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/7/11/560/s1. Figure S1. Physical map of the JTM locus in the NEA12 and E1 genomes. The JTM locus has 13 predicted and known genes in four clusters. Cluster 1 (ltmG, ltmS, ltmM, ltmK), Cluster 2 (ltmP, ltmQ, ltmF, ltmC, ltmB), Cluster 3 (jtmD and jtmO) and Cluster 4 (jtm01, jtm02). Light grey arrows display predicted and known genes and their orientation. The locations of the pks pseudogene, transposase with a MULE domain (PP03) and Helitron helicase-like transposable element (TE) are also shown. PP = predicted protein; TE = transposable element; ψ = pseudogene. Figure S2. Trees generated through ML analysis of the predicted amino acid sequence of JTM locus genes from NEA12 and E1 and selected top BLASTp hits in the NCBI database. Genbank accession number and percent amino acid identity created by Clustal2.1, and species name is provided for each protein.