Metabolomic-guided discovery of cyclic nonribosomal peptides from Xylaria ellisii sp. nov., a leaf and stem endophyte of Vaccinium angustifolium

Fungal endophytes are sources of novel bioactive compounds but relatively few agriculturally important fruiting plants harboring endophytes have been carefully studied. Previously, we identified a griseofulvin-producing Xylaria species isolated from Vaccinium angustifolium, V. corymbosum, and Pinus strobus. Morphological and genomic analysis determined that it was a new species, described here as Xylaria ellisii. Untargeted high-resolution LC-MS metabolomic analysis of the extracted filtrates and mycelium from 15 blueberry isolates of this endophyte revealed differences in their metabolite profiles. Toxicity screening of the extracts showed that bioactivity was not linked to production of griseofulvin, indicating this species was making additional bioactive compounds. Multivariate statistical analysis of LC-MS data was used to identify key outlier features in the spectra. This allowed potentially new compounds to be targeted for isolation and characterization. This approach resulted in the discovery of eight new proline-containing cyclic nonribosomal peptides, which we have given the trivial names ellisiiamides A-H. Three of these peptides were purified and their structures elucidated by one and two-dimensional nuclear magnetic resonance spectroscopy (1D and 2D NMR) and high-resolution tandem mass spectrometry (HRMS/MS) analysis. The remaining five new compounds were identified and annotated by high-resolution mass spectrometry. Ellisiiamide A demonstrated Gram-negative activity against Escherichia coli BW25113, which is the first reported for this scaffold. Additionally, several known natural products including griseofulvin, dechlorogriseofulvin, epoxy/cytochalasin D, zygosporin E, hirsutatin A, cyclic pentapeptides #1–2 and xylariotide A were also characterized from this species.


Results
Identification, biology and ecology of Xylaria sp. Approximately 30 strains of Xylaria sp. were isolated from surface-sterilized blueberry tissues collected from highbush and wild blueberry fields within a ~300 × 100 km triangular area. All fields were surrounded by forested lands. Preliminary phylogenetic analysis using the internal transcribed spacer (ITS) barcode combined with morphological features confirmed conspecificity of isolated endophytic Xylaria sp. strains. However, identification of the strains to species was not possible using molecular or in vitro morphological data. Based on a BLAST query of the Xylaria sp. ITS and RPB2 sequences with available GenBank sequences, the endophyte strains were closest related to sequences identified as Xylaria berteri, X. castorea, X. cubensis, X. laevis, and X. longipes, species that form conspicuous sexual reproductive structures (stromata) from decaying hardwood. Given the close phylogenetic relationship of the unknown Xylaria endophyte to these species and evidence of prevalent endophytic-saprotrophic life histories within Xylariaceae [23][24][25]41,42 , we inferred that the unknown Xylaria endophyte likely produces stromata from decaying hardwood in mixedwood stands in the Acadian forest. Thus, Xylaria stromata were selectively sampled during ongoing field surveys to collect the putative sexual state of the endophyte. This would provide material for identification and insight into its life history 20,43 . Lc-MS analysis of culture extracts and multivariate data analysis. Fifteen strains of X. ellisii were subject to further study: four from cultivated highbush blueberry plants and 11 from wild blueberry plants. Ethyl acetate extracts of the culture filtrate and associated mycelium were screened using standardized LC-UV/MS conditions.
In order to identify unique secondary metabolite differences between extracts of Xylaria isolates of highbush and wild blueberry plants we compared the extracted filtrates and mycelium with three different pair-wise comparisons. These comparisons included: ethyl acetate extracts of Xylaria strains grown on 2% malt extract broth (ML) versus those grown in potato dextrose broth (PDB) cultures; ML media cultures of highbush versus wild varieties; and, PDB medium cultures of highbush versus wild isolates (Fig. 1S). A supervised multivariant analysis method, Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA), was used to identify outlier metabolites biosynthesized under the different culture conditions tested. OPLS-DA correlates differences in secondary metabolite feature abundances (X variables) to various treatment groups (Y variables) by identifying principle components that describe differences. R 2 X, R 2 Y, and Q 2 parameters are important validation parameters used for OPLS-DA, where R 2 X and R 2 Y describes the percentage of X and Y variables described by the model (Fig. 4 and Supplementary Fig. 2S). A valid model is defined as having a prediction statistic of Q 2 > 0.4, with values above 0.7 being highly significant 46 . Metabolite features with a high Variable Importance in Projection (VIP) scores (>0. 7) are responsible for driving the differences between treatment groups, and these values are considered significant 47 . Their metabolic features can be viewed at both ends of the OPLS-DA S-plot.

Discussion
Xylaria ellisii was the most commonly isolated Xylariaceae endophyte from Picea and Pinus in Eastern Canada 57 . Stromata of X. ellisii were commonly found on decaying Acer saccharum branches or stems in the same forest stands where it was isolated as a Picea endophyte. Endophyte ITS sequences in GenBank corresponding to X. ellisii originate from an exceptional diversity of hosts, including Tsuga canadensis, bryophytes (e.g.: Hypnum sp.), liverworts (e.g.: Metzgeria furcata, Trichocolea tomentella), and lichens (e.g.: Flavoparmelia caperata, Sticta beauvoisii, Xanthoparmelia conspersa) (Fig. 3). In New Brunswick, corresponding X. ellisii stromata were commonly found in late summer and early fall only on decaying Acer saccharum wood; however, the stromatal host range is likely broad. For example, Laessøe (1987)  Xylaria ellisii is a common Picea and Pinus endophyte even in conifer-dominated stands lacking Acer saccharum or any other hardwood hosts possibly suitable for the production of stromata. This indicates that the fungus is capable of persisting in the environment in the prolonged absence of a suitable primary host. The method of transmission between foliage is currently unknown. It is conceivable that the dry, powdery masses of conidia produced from conidiomata in vitro are also produced on dead foliage and capable of infecting new foliage by means of air currents or insect vectors 58,59 . Abscised foliage infected with X. ellisii is probably capable of saprotrophically colonizing hosts by means of direct contact (viaphytism), as demonstrated in other Xylaria species 25 . The known range of hosts that X. ellisii can endophytically infect includes lichens and various understory and overstory plant species with different successional statuses, allowing for its persistence across forest succession pathways and disturbances (e.g.: as an endophyte of the fire-adapted seral species Vaccinium angustifolium). A proposed endophytic-saprotrophic life history is described and illustrated for Xylaria ellisii (as Xylaria sp.) by Tanney et al. 60 .
The production of the potently antifungal compound griseofulvin by X. ellisii, an apparently ubiquitous endophyte with a broad host range, is significant. Griseofulvin is toxic to a wide variety of plant pathogens [61][62][63][64] and is systemically translocated within plants 65 , suggesting that X. ellisii endophyte infections could increase host resistance to plant pathogens. For example, Park et al. (2005) described griseofulvin production in an unidentified Xylaria endophyte of Abies holophylla and showed its ability to control the development of plant diseases such as barley powdery mildew (Blumeria graminis f. sp. hordei), rice sheath blight (Corticium sasaki), wheat leaf rust (Puccinia recondita), and rice blast (Magnaporthe grisea) 64 . Griseofulvin and related compounds are reported from Xylaria endophytes of Asimina triloba, Chrysobalanus icaco, and Garcinia hombroniana [66][67][68] . Richardson et al. (2014) reported the production of the antifungal compound griseofulvin by Xylaria ellisii (as Xylaria sp.) isolated as a foliar endophyte of Pinus strobus and Vaccinium angustifolium 38 . These isolates produced griseofulvin and its de-halogenated analogue (Fig. 1), along with piliformic acid 38 . Subsequent investigations of white pine seedlings infected with this Xylaria species found griseofulvin at biologically effective concentrations in the needles 69 .
In this study, we have applied a LC-MS metabolomic guided discovery approach to profile the chemical space of a novel endophytic species described here as Xylaria ellisii. Our collections of isolates have identical ITS DNA sequences yet differ in their LC-MS metabolite profiles and bioactivity. OPLS-DA and S-plot analysis identified features separated by a statistical toll, Variable Importance in Projection (VIP) scores. VIP scores from the extracted filtrates and mycelium extracts were calculated and extracts differentiated by this method were targeted for compound isolation and structural characterization. This approach resulted in the discovery of three new cyclic pentapeptides given the trivial names ellisiiamides A-C (12)(13)(14) and the putative identification and annotation of ellisiiamides D-H by LC-HRMS and LC-HRMS/MS analysis. Additionally, 11 known compounds are reported to be produced by these strains. Ellisiiamide A (12) was active against Gram-negative bacteria and is a first report for this scaffold. These findings are of interest as the isolates were also reported from eastern white pine needles in a pine-blueberry forest ecotype. Endophytes from wild Vaccinium species may be an interesting source of novel bioactive compounds. This information provides a better understanding of the chemical ecology of plant-fungi microbiomes. In the long term, opportunities may present to employ this information for integrated pest management crop protection strategies.

Methods
Sampling, isolation, and culturing. Plant material, including leaves and stems from highbush and wild blueberries, were collected from three different locations within the Acadian forest region of Nova Scotia, Canada. Highbush blueberry endophyte isolates were obtained from a commercial field in Rawdon, Nova Scotia and wild blueberry endophytes isolates were collected from commercial fields in Mount Thom, Debert, and Portapique, Nova Scotia. Specimens were collected in labelled bags and stored at -20 °C prior to fungal isolation. Plant tissues were first washed with sterile deionized water to remove any loose debris and surface contaminants, followed by a chemical surface-sterilization process using sodium hypochlorite bleach (6%) and ethanol (70%). Small segments were then cut and/or incised and placed in Petri plates containing 2% malt extract agar (MEA; 20 g Bacto malt extract, Difco Laboratories, Sparks, USA; 15 g agar, EMD Chemicals Inc., Gibbstown, USA; 1 L deionized H 2 O). Inoculated plates were incubated at 25 °C for 4-8 weeks, depending on the presence of filamentous hyphae. Endophytic fungi that grew from cut ends were then transferred to potato dextrose agar (PDA, Sigma-Aldrich, Canada) plates and incubated at 25 °C.
Field specimens of stromata were collected and stored in paper bags. Single-ascospore isolates were made by affixing with petroleum jelly a small (ca. 5 mm 2 ) piece of stroma containing mature perithecia to the lid of a Petri dish containing water agar (WA; 15 g agar, EMD Chemicals Inc., Gibbstown, USA; 1 L deionized H 2 O). Germination of ejected ascospores on the agar surface was confirmed by stereo microscope (Olympus SZX12, Olympus, Tokyo, Japan) and germinating ascospores were transferred to individual Petri plates containing 2% MEA and incubated at 20 °C. Dried specimens were accessioned in the Canadian National Mycological Herbarium (Ottawa, Ont.; DAOM). Living cultures were deposited in the Canadian Collection of Fungal Cultures (Ottawa, Ont.; DAOMC). Additional specimens used for morphological comparison and phylogenetic analyses were also obtained from DAOM, DAOMC, and the personal culture collection of J.B. Tanney.
Xylaria strains from highbush blueberry and wild blueberry were cultured in PDB (24 g/L potato dextrose broth) and ML (30 g/L malt) fermentation media. Each strain was grown in 1 L Roux bottles containing 200 mL of media and grown statically for 4-6 weeks at 25 °C. The culture broth was then separated from the mycelium by vacuum filtration using a Whatman #4 filter paper. The filtrate was extracted with equal volumes of ethyl acetate, while the mycelium was first lyophilized for 24 h and then extracted with equivalent volumes of methanol and acetone (1:1). Organic fractions were then dried under reduced pressure by rotary vacuum. Extracts were then re-suspended in 600 μL of HPLC grade acetonitrile with minimal amounts of DMSO added for solubility. The filtrates were then centrifuged at 13,000 rpm for 15 min and Acro-disk (13 mm, 0.45 μm GHP) filtered prior to LC-MS analysis.
Morphological study. Sections of stromata were cut by hand using a safety razor blade or with a freezing microtome (ca. 15-30 µm thick) and mounted in either water, 5% KOH, 85% lactic acid, or Lugol's solution with or without 5% KOH pretreatment to test amyloid reactions 74 . Stromata and colony colours were described using alphanumeric codes 75 . Observations of the asexual morph were made from living cultures grown on oatmeal agar (OA) 76 . Microscopic measurements were taken from living material mounted in deionized water and are presented as ranges calculated from the mean ± standard deviation of each measured value with outliers in brackets. Observations were made using an Olympus BX50F4 light microscope and an Olympus SZX12 stereo microscope (Olympus, Tokyo, Japan). Images were captured with an InfinityX-32 camera (Lumenera Corp., Ottawa, Canada) using Infinity Analyze v. 6 Loci chosen for sequencing included the internal transcribed spacer rDNA region (ITS), β-tubulin (BenA), translation elongation factor 1-alpha (EF1-α), the second largest subunit of RNA polymerase II (RPB2), 18 s nuc rDNA (SSU), and 28 S nuc rDNA (LSU). Primer pairs used for PCR amplification and sequencing included: ITS1 and ITS4 77 or ITS4A and ITS5 78 for ITS; Bt2a and Bt2b for BenA 79 ; RPB2-5f2 and RPB2-7CR 80 for RPB2; and EF1-728F and EF1-986R 81 for EF1-α. LSU was amplified using LR0R and LR5 and sequenced using the primers LR0R, LR3, LR3R, and LR5 82 . SSU was amplified using the primers NS1 and NS4, and sequenced using the primers NS1, NS2, NS3, and NS4 77 . PCR and sequencing were performed as described by Tanney and Seifert (2017) 57 . To improve ITS amplification in herbarium specimens, 0.5 μm of 20 mg/ml bovine serum albumin (BSA) was added per reaction.
For all analyses, sequences were aligned using MAFFT v7 83 and visually inspected and manually aligned when necessary in Geneious R8 v8.1.5 (Biomatters, Auckland, New Zealand). The most suitable sequence evolution model was determined based on the optimal Akaike information criterion scores in MrModeltest v2.2.6 84 . Consensus trees were visualized in FigTree 1.4.2 (available at http://tree.bio.ed.ac.uk/software/figtree/) and exported as SVG vector graphics for assembly in Adobe Illustrator v10 (Adobe Systems, San Jose, CA, USA).
Three separate phylogenetic analyses were performed. The first phylogeny included ITS sequences of diverse representative endophytes isolated from highbush and wild blueberry leaves and stem. The ex-type of Mucor ellipsoideus (ATCC MYA-4767; NR_111683) was selected as outgroup because of its basal position (Mucoromycotina). Maximum likelihood (ML) analysis was performed using RAxML v8.2.4 in PAUP v4.0b10 starting from a random starting tree with 1000 bootstrap replicates 85,86 .
The second phylogenetic analysis included RPB2 sequences from related Xylaria species. The resulting alignment was 1058 bp long and consisted of 47 taxa, including the outgroup Barrmaelia rhamnicola (CBS 142772). Bayesian analysis was performed using MrBayes v3.2.6 87 . Three independent Markov Chain Monte Carlo (MCMC) samplings were performed with 12 chains (11 heated and one cold) with sampling every 500 generations until the standard deviation of split frequencies was <0.01. The first 25% of trees were discarded as burn-in and the remaining trees were kept and combined into one consensus tree with 50% majority rule consensus. Convergence was assessed from the three independent runs using Tracer v1.6 88 . The third phylogenetic analysis included ITS sequences from related endophytic Xylaria isolates. The alignment was 593 bp long and included sequences from 107 isolates or samples. The resulting phylogenetic analysis was performed in the same manner as described above, with Nemania serpens (GU292820) as the outgroup.
All novel sequences used in this study were accessioned in GenBank (Supplementary Table 13S) and taxonomic novelties and associated metadata were deposited in MycoBank (www.MycoBank.org).

Lc-UV/HRMS and Lc-UV/HRMS/MS screening.
Extracts of endophytic cultures were screened using a Dionex Ultimate 3000 HPLC-UV system coupled to a Bruker maXis 4 G ultra-high-resolution-qTOF mass spectrometer operated in positive electrospray ionization (ESI) with calibrations done using sodium formate ion clusters. LC-MS data were collected using a scan range of 150-1100 m/z, with the nebulizer gas (nitrogen) at 3 bar, dry gas flow at 8 L/min, dry gas temperature at 240 °C, and capillary voltage at 4500 V. Chromatographic separations were performed using a standardized HPLC-UV method with a Supelco Ascentis Express C18 reverse-phase core-shell column (150 × 4.6 mm, 2.7 μm, Sigma Aldrich, USA) operating at 750 μL/min and at 40 °C. UV/vis data were acquired from 190-600 nm and monitored at four wavelengths (210, 254, 275 and 350 nm). Mobile phase composition was linear with a gradient of 5% organic from 0 to 1 min, 5-95% from 1 to 24 min, 95-100% from 24 to 25 min, and 100% from 25 to 31 min. Solvent A was H 2 O + 0.1% formic acid and solvent B was acetonitrile with 0.1% formic acid (v/v). HR-MS/MS analysis was performed on a Thermo Q-Exactive Orbitrap mass spectrometer operated in positive electrospray ionization (ESI+) and coupled to an Agilent 1290 HPLC system. Data processing and multivariate statistical analysis. Data processing and analyses were modified from a previously published protocol (Fei et al., 2014). Post-acquisition internal calibration using sodium formate clusters in both ESI+ and ESI-were performed with Bruker's Data Analysis 4.0 SP4. LC-MS data files were converted to.mzXML format using Bruker CompassXport. Metabolic features were extracted and aligned using open source XCMS with centWave algorithm 89 ; adducts, isotopic ions, and in-source fragments were identified using CAMERA 90,91 . To acquire the final metabolite feature list, isotopic ions and features with integrated peak area under 10,000 were removed. For mycelium metabolome, metabolite features that eluted after 25 min were eliminated.
Both extracted filtrates and mycelium were analyzed using supervised multivariate OPLS-DA after pareto scaling by SIMCA-P+ 12.0.1 (Umetrics, Kinnelon, NJ). The statistical parameters R 2 X(cum), R 2 Y(cum), and Q 2 (cum) of OPLS-DA were used to assess the fitness of the model. R 2 X (R2Y) indicated the fraction in which metabolite features (X) and group (Y) matrix was were explained by the model. A prediction statistic (Q 2 ) above 0.4 was indicative of a statistically robust model, i.e. true differences between the comparing groups, and Q2 between 0.7-1.0 indicated the model was statistically significant 46 . Both R 2 and Q 2 followed an upward trend from 0 to 1. For an over-fit model, R 2 approached 1, and Q 2 fell toward 0 92 . Significant features between classes were identified based on OPLS-DA S-plot and their Variable Importance in Projection (VIP) score. To ensure the identified metabolites are the sole important markers, the two OPLS-DA analyses were conducted in parallel by only including the significant features or by removing the significant features from the raw data 92 . A useful metabolite subset was produced if the first model was successful and the later model failed. (2020) 10:4599 | https://doi.org/10.1038/s41598-020-61088-x www.nature.com/scientificreports www.nature.com/scientificreports/ Metabolite isolation and characterization. NMR experiments for 1D and 2D measurements were performed on a Bruker Advance III 700 MHz NMR spectrometer equipped with a 5 mm QNP cryoprobe, operating at 700.17 MHz for 1 H NMR and 176.08 MHz for 13 C NMR or a Bruker Advance III HD 850 MhZ NMR spectrometer equipped with a 5 mm TXI probe operating at 850.21 MHz for 1 H NMR and 213.81 MHz for 13 C NMR, with chemical shifts referenced to the residual solvent signal 93 . Nitrogen dried compounds were re-suspended in 200 μL of deuterated solvent (C6D6, CD3OD, or DMSO-d6) and transferred to 3 mm NMR tubes (Wilmad 335-pp-7) for NMR measurements. NMR data processing was done using MNOVA NMR software ver. 10.0.1 by Mestrelab Research. Optical rotation measurements were done using an Autopol IV Polarimeter (Rudolph Research Analytical).
Purification of metabolomic targeted metabolites was performed on a semi-preparative HPLC system consisting of an Agilent 1100 series HPLC with a G1311A Quaternary Pump, a G1379A Degasser, a G1367A Wellplate Autosampler, a G1316A Column Thermostat, a G1315B Diode Array Detector (DAD), and a G1364C Automatic Fraction Collector controlled by Agilent ChemStation software (Rev. B.03.02-SR2). Metabolites were isolated using a Phenomenex Synergi-Max reverse-phase C-12 column (250 × 10 mm, 4 μm) (Torrence, CA, USA) operating at 5 mL/min and 40 °C. Mobile phase composition was a linear gradient of 5% organic from 0 to 3 min, 5-30% from 3 to 16 min, 30% from 16 to 20 min, and 30-85% from 20-37 min with fractions collected every 20 s. Known isolated compounds (mg/L): dechlorogriseofulvin (2)   Biological activity screening. Compounds were tested for their minimum inhibitory concentration (MIC) according to the Clinical Laboratory Standards Institute (CLSI) protocols M7-A5 and M27-A (National Committee for Clinical Laboratory Standards, 2000, 1997). Stock working solutions were made to 5, 10, and 20 mg/mL and tested at a maximum concentration of 200 μg/mL in 96-well liquid culture (National Committee for Clinical Laboratory Standards, 1997, 2003) as previously described 37 . Preliminary evaluation of biological activity was against E. coli BW25113 ΔbamBΔtolC, a membrane and efflux pump compromised strain, Staphylococcus aureus ATCC# 29213, Bacillus subtilis 1A1, Micrococcus luteus, Saccharomyces cerevisiae B4741, and Candida albicans ATCC# 90028. A cut-off of <25% growth was used for inhibition, with the trend across dilutions also considered 37 .