Cyclodepsipeptide Biosynthesis in Hypocreales Fungi and Sequence Divergence of The Non-Ribosomal Peptide Synthase Genes

Fungi from the Hypocreales order synthesize a range of toxic non-ribosomal cyclic peptides with antimicrobial, insecticidal and cytotoxic activities. Entomopathogenic Beauveria, Isaria and Cordyceps as well as phytopathogenic Fusarium spp. are known producers of beauvericins (BEAs), beauvenniatins (BEAEs) or enniatins (ENNs). The compounds are synthesized by beauvericin/enniatin synthase (BEAS/ESYN1), which shows significant sequence divergence among Hypocreales members. We investigated ENN, BEA and BEAE production among entomopathogenic (Beauveria, Cordyceps, Isaria) and phytopathogenic (Fusarium) fungi; BEA and ENNs were quantified using an LC-MS/MS method. Phylogenetic analysis of partial sequences of putative BEAS/ESYN1 amplicons was also made. Nineteen fungal strains were identified based on sequence analysis of amplified ITS and tef-1α regions. BEA was produced by all investigated fungi, with F. proliferatum and F. concentricum being the most efficient producers. ENNs were synthesized mostly by F. acuminatum, F. avenaceum and C. confragosa. The phylogeny reconstruction suggests that ancestral BEA biosynthesis independently diverged into biosynthesis of other compounds. The divergent positioning of three Fusarium isolates raises the possibility of parallel acquisition of cyclic depsipeptide synthases in ancient complexes within Fusarium genus. Different fungi have independently evolved NRPS genes involved in depsipeptide biosynthesis, with functional adaptation towards biosynthesis of overlapping yet diversified metabolite profiles.


Introduction
The higher fungi (Dikarya) constitute one of the largest groups of microbial eukaryotes, responsible for producing numerous bioactive secondary metabolites. A diverse set of fungi belonging to the Hypocreales order synthesize non-ribosomal cyclic peptides [1,2]. Some of these have been recognized as emerging toxins and agents of competition between divergent taxa. In particular, the entomopathogenic genera Beauveria, Isaria and Cordyceps as well as various phytopathogenic Fusarium spp. are known and phylogenetic analyses allowed us to pinpoint novel strains of cyclodepsipeptide producers and place their chemotypes into the context of extant genetic diversity.

Fungal Species Identification
Nineteen fungal strains were identified and used in the study. Some of the analyzed strains represented a group of plant pathogens (nine Fusarium strains belonging to six species), often isolated from such crop species as wheat, rye, pea or asparagus [42]. The second group of identified fungi consisted of selected strains representing the well-known entomopathogenic genera. Among these, representatives of Beauveria, Cordyceps and Isaria genera were present. All the above genera are comprised of species largely characterized as the broad host range facultative entomopathogens, important for control of insect populations in nature and with a possible use in biocontrol [43][44][45]. As such, they are frequently considered as a possible microbial insecticide against insects feeding on crops, including lepidopterans and coleopterans [46][47][48].
The molecular identification of all fungal strains was performed using sequence analysis of the obtained amplicons from polymerase chain reaction (PCR). The DNA regions were amplified with ITS4/ITS5 primers (for entomopathogens) and Ef728M/TefR1 primers (for phytopathogens) and were subsequently sequenced. To confirm species identification, amplified DNA fragments were compared with reference genes from the GenBank Database.
Acquired sequences by the amplification of ITS and tef -1α fragments from the investigated fungal strains showed over 98% similarity to the reference genes. The high identity level to the reference sequences from GenBank showed that all fungal species have been identified correctly. Results of molecular identification and characterization of fungal strains are described in Table 1. Relationships between examined isolates were ascertained as maximum likelihood phylogenetic tree reconstructions based on the ITS1 and tef -1α sequences. For phylogeny of the internal transcribed spacer (Figure 2 The analyses of the ITS and tef -1α genes have been frequently employed in the phylogenetic studies of fungi belonging to the Hypocreales order [3,[49][50][51][52][53]. However, based only on these conserved taxonomic markers, the determination of differences between closely related genotypes is often limited. The analysis of secondary metabolite biosynthesis-related genes' sequences, revealing the intraspecific and interspecific polymorphisms in datasets comprised of isolates from multiple species/genera, is often a better solution [5,42,54]. 1 Reported from precursor-directed biosynthesis. 2 Detected and tentatively characterized in this study. 3 Detected and quantified in this study. * Unreported BEAE

Non-Ribosomal Peptide Synthetase Genes Divergence
The phylogeny reconstructions of housekeeping sequences (ITS- Figure 2, tef -1α- Figure 3) point to the ancestral biosynthesis of beauvericin independently diverging into biosynthesis of other compounds (enniatins, beauvenniatins, allobeauvericins) across different taxa. The evolutionary relationships suggested by beauvericin gene fragments seem to support the notion of B. felina as an early diverging beauvericin producer, however, the obtained sequence fragment closely matches the cyclosporin synthase recently elucidated by Xu et al. [55], so verification by cloning, functional genomics and/or whole genome sequencing is likely needed to fully verify the synthase sequence. The divergent positioning of three Fusarium isolates (1337, 41.9.3, P36) as well as an unusual GenBank reference sequence for F. oxysporum f.sp. cucumerinum (itself highly similar to the known enniatin synthase from F. scirpi, accession Z18755) raises the possibility of parallel acquisition of cyclic depsipeptide synthases in ancient complexes within Fusarium genus.
Notably, such separation of two diverging variants of the synthase was previously observed in our earlier parsimony-based reconstruction [42], where predominantly enniatin-producing F. avenaceum/acuminatum/tricinctum isolates clustered distinctly separate from beauvericin-producing strains. Regardless of this, save for the B. felina sequence, all Cordycipitaceae isolate sequences form a well-supported clade with distinct separation between bassianolide synthase-like sequences (canonical bassianolide synthase from B. bassiana, model sequences from C. confragosa and C. brongniartii, isolates 4414 and ENC4) and beauvericin synthase-like fragments (canonical beauvericin synthase from B. bassiana, sequences from isolates tentatively identified as I. farinosa, I. fumosorosea, as well as single C. confragosa isolate ENC6). It is worth noting that the fairly recent update of Cordycipitaceae nomenclature and phylogeny shows that past identification of even reference Isaria/Cordyceps strains available from well-curated collections was not well supported by newly available molecular data [56]. In the context of continued research in the field [57], I. farinosa in particular is demonstrated to be a wide label spanning different species within the redefined, monophyletic Cordyceps and Akanthomyces genera ( Figure 4).
Pathogens 2020, 9, x FOR PEER REVIEW 2 of 22 Figure 3. Phylogeny of 16 tef-1α partial sequences. Numbers below branches indicate support, box outlines mark reference sequences, and annotated compound classes are listed beside the leaves. The majority-rule consensus tree was based on the alignment of 435 columns (including 87 coding triplets) with 101 parsimony informative sites. The following models were chosen by IQTREE based on BIC criterion: TN + F + G4 for intronic fragments and MG + F3X4 + I for coding parts. A total of 102 ultrafast bootstrap iterations were carried out (automated stopping criterion for tree search). All bipartitions with below 50% ultrafast bootstrap support were collapsed in the final representation.

Non-Ribosomal Peptide Synthetase Genes Divergence
The phylogeny reconstructions of housekeeping sequences (ITS- Figure 2, tef-1α- Figure 3) point to the ancestral biosynthesis of beauvericin independently diverging into biosynthesis of other compounds (enniatins, beauvenniatins, allobeauvericins) across different taxa. The evolutionary relationships suggested by beauvericin gene fragments seem to support the notion of B. felina as an early diverging beauvericin producer, however, the obtained sequence fragment closely matches the cyclosporin synthase recently elucidated by Xu et al. [55], so verification by cloning, functional genomics and/or whole genome sequencing is likely needed to fully verify the synthase sequence. The divergent positioning of three Fusarium isolates (1337, 41.9.3, P36) as well as an unusual GenBank reference sequence for F. oxysporum f.sp. cucumerinum (itself highly similar to the known enniatin synthase from F. scirpi, accession Z18755) raises the possibility of parallel acquisition of cyclic depsipeptide synthases in ancient complexes within Fusarium genus.
Notably, such separation of two diverging variants of the synthase was previously observed in our earlier parsimony-based reconstruction [42], where predominantly enniatin-producing F. avenaceum/acuminatum/tricinctum isolates clustered distinctly separate from beauvericin-producing strains. Regardless of this, save for the B. felina sequence, all Cordycipitaceae isolate sequences form a well-supported clade with distinct separation between bassianolide synthase-like sequences (canonical bassianolide synthase from B. bassiana, model sequences from C. confragosa and C. brongniartii, isolates 4414 and ENC4) and beauvericin synthase-like fragments (canonical beauvericin synthase from B. bassiana, sequences from isolates tentatively identified as I. farinosa, I. fumosorosea, as well as single C. confragosa isolate ENC6). It is worth noting that the fairly recent update of Figure 3. Phylogeny of 16 tef -1α partial sequences. Numbers below branches indicate support, box outlines mark reference sequences, and annotated compound classes are listed beside the leaves. The majority-rule consensus tree was based on the alignment of 435 columns (including 87 coding triplets) with 101 parsimony informative sites. The following models were chosen by IQTREE based on BIC criterion: TN + F + G4 for intronic fragments and MG + F3X4 + I for coding parts. A total of 102 ultrafast bootstrap iterations were carried out (automated stopping criterion for tree search). All bipartitions with below 50% ultrafast bootstrap support were collapsed in the final representation.
With regards to the sequencing and phylogeny of putative BEAS homolog fragments, the validation of these sequences on the level of full genomic sequence and functional experiments is a desired follow-up. In particular, the further study of the ENC3 Beauveria felina isolate on genomic and functional levels is likely to provide insight as to the diversification of beauvericin/beauvenniatin/enniatin biosynthesis across this important genus of insect pathogens ( Figure 4).

In Vitro Cyclodepsipeptide Biosynthesis
Quantitative analysis of ENN A (compound 12 in Table 1), ENN A 1 (13), ENN B (14), ENN B 1 (15) and BEA (1) was conducted using LC-MS/MS (Table 2). Beauvericin was produced by all investigated fungi, with the most efficient producers being Fusarium strains, such as F. proliferatum (RT6.7, RT 5.4) and F. concentricum (P35). To some extent, all of the studied fungi are known producers of beauvericin [42], however, it is important to note that each fungal strain possesses unique abilities to produce different amounts of BEA, which can also be affected by the cultivation conditions [58,59].
In contrast, the most efficient producers of the four enniatin analogues were F. acuminatum (41/9/3) and F. avenaceum (1337). The production of enniatins by phytopathogenic fungi belonging to the genus Fusarium has been described earlier [5][6][7]17,[28][29][30]42]. The most efficient enniatin producer among entomopathogenic fungi was the isolate of C. confragosa (4414). Notably, some of the investigated fungal strains were not able to biosynthesize any of the four enniatins, and this group included entomopathogenic fungi B. bassiana (MU3), B. felina (ENC3), I. farinosa (4447, ENC5, ENC9, MU5) and C. confragosa (ENC1, ENC6). The same entomopathogenic fungal species were previously only investigated for beauvericin production, or production of beauvericin analogues [9,[11][12][13]. Indeed, we were not able to find any previous reports concerning the production of enniatins by those entomopathogenic fungi. On the contrary, our own analyses showed that for at least two strains-I. fumosorosea (MU1) and C. confragosa (4414)-a limited capacity to produce enniatins exists, as each isolate was able to synthesize two of the four enniatin analogues (Table 2). Moreover, the two phytopathogenic fungal species F. proliferatum and F. oxysporum did not synthesize enniatin analogues, although they have earlier been described as enniatin producers [16,29,42,60]. This suggests a need for comprehensive reevaluation of cyclodepsipeptide profiles across multiple representatives of known producing taxa, in particular as the production of enniatins may depend on cultivation conditions, storage conditions or the host and the environment from which the fungi were isolated [59].    Liquid chromatography-ion-trap mass spectrometry (LC-ITMS) was used for the analysis of BEAs and BEAEs appearance in all investigated fungal crude extracts (Table 3). Two selected strains (P35 and 4447) were also analyzed using liquid chromatography-high-resolution mass spectrometry (LC-HRMS) for accurate mass measurements and HRMS/MS (Table 4). We have previously tentatively shown that these two strains produce BEA A, BEA C, BEA D, BEA F, BEA J, BEA K, BEAE A, and BEAE L [3]. Only five of the studied strains (P35, RT 6.7, 4447, MAL 1.4, and A4.12) were found to produce a wide range of different BEA and BEAE analogues (>15) (Table 3, Figure 1).
Using reversed-phase liquid chromatography (RPLC), the elution of individual BEAs and BEAEs was highly related to their molecular weights, i.e., beauvericin analogues with a higher molecular weight eluted later ( Table 4). The presumed BEAs and BEAEs provided plentiful [M + NH 4 ] + ions during electrospray ionization, while [M + H] + and [M + Na] + ions accounted for only few percent of the total ion yield. In total, five beauvericins (compounds 1-5), five beauvenniatins (6-10) and one allobeauvericin (11) were tentatively detected in different fungal extracts (Figures 1 and 5, and Table 4).        BEA itself was the most common analogue and was produced in the highest relative amounts (in terms of LC-MS peak areas) by all studied fungal strains. We used BEA (1) as a reference for comparing product-ion spectra of the sodiated ions of different analogues [3]. All BEA analogues, characterized by the presence of N-methyl-phenylalanine in their structure, were tentatively identified by the presence of a product-ion due to a loss of −161 Da. The beauvericins that were observed in the studied strains varied in hydroxy acid composition (Figure 6a,b). BEA E (3) contains one leucine moiety in addition to two N-methyl-phenylalanine units, and this analogue was tentatively identified by the presence of a product-ion corresponding to a loss of −127 Da. BEA E (3) possesses three D-Hiv units giving rise to product-ions corresponding to a loss of −100 Da (Figure 6b) [11,24]. In contrast, BEA B (2) possesses two D-Hmp units and one D-Hiv unit, giving rise to product-ions corresponding to −114 and −100 Da, respectively (Figure 6a) [12]. Furthermore, BEA G 1 , and G 2 (4, 5) consist of one or two D-Hbu groups, respectively, which can be observed by product-ions, corresponding to a loss of −86 Da (Figure 7a,b) [13].  ALLOBEAs A, B (11) and C are isomers of BEAs A/F, B (2) and C, which are composed of three N-methyl-phenylalanine moieties and one, two or three D-Hmp groups, respectively [12]. BEAEs 7-9, which are also isomers [10], were produced in the smallest relative amounts. The HRMS/MS spectra of their sodiated ions were of poor signal/noise ratios; still, they were better than spectra of their ammoniated ions. The detected isomers of depsipeptides consist of the same moieties, therefore, we are not able to tell which isomer/s is/are present in the sample ( Figure 5). The differences between studied compounds (2/11 and 7/8/9) were described earlier, on the basis of NMR analyses [10,12]. All investigated BEAEs (6-10) contain one group of N-methyl-valine and two N-methyl-phenylalanine residues, which was manifested by product-ions corresponding to losses of −161 and −113 Da (Figure 8a-c). BEAE B (6) contains three D-Hiv groups (Figure 8a) [10], while BEAEs 7-9 (Figure 8b) appeared to contain one D-Hmp group instead of D-Hiv, based on our ALLOBEAs A, B (11) and C are isomers of BEAs A/F, B (2) and C, which are composed of three N-methyl-phenylalanine moieties and one, two or three D-Hmp groups, respectively [12]. BEAEs 7-9, which are also isomers [10], were produced in the smallest relative amounts. The HRMS/MS spectra of their sodiated ions were of poor signal/noise ratios; still, they were better than spectra of their ammoniated ions. The detected isomers of depsipeptides consist of the same moieties, therefore, we are not able to tell which isomer/s is/are present in the sample ( Figure 5). The differences between studied compounds (2/11 and 7/8/9) were described earlier, on the basis of NMR analyses [10,12]. All investigated BEAEs (6-10) contain one group of N-methyl-valine and two N-methyl-phenylalanine residues, which was manifested by product-ions corresponding to losses of −161 and −113 Da (Figure 8a-c). BEAE B (6) contains three D-Hiv groups (Figure 8a) [10], while BEAEs 7-9 (Figure 8b) appeared to contain one D-Hmp group instead of D-Hiv, based on our HRMS/MS data. Moreover, we detected one compound (10) in the crude extract from F. concentricum (P35) pure rice cultures that appeared to be composed of an unreported combination of amino and hydroxy acids. Our HRMS/MS data showed that it could be an unreported BEAE consisting of three D-Hiv groups, one N-methyl-valine, N-methyl-phenylalanine and N-methyl-leucine or -isoleucine group (Figure 8c). The latter two cannot be distinguished by mass spectrometry alone. For future investigations, nuclear magnetic resonance (NMR) will be considered in order to better discern the actual depsipeptide structure.
Pathogens 2020, 9, x FOR PEER REVIEW 2 of 22 investigations, nuclear magnetic resonance (NMR) will be considered in order to better discern the actual depsipeptide structure.

Fungal Strains, Media and Growth Conditions
Fungal strains from the Hypocreales order were isolated from infected plants and insects found in Poland. All isolated strains were placed in the fungi collection of the Institute of Plant Genetics, Polish Academy of Sciences, Poznań, Poland. For characterization of fungal isolates, the species, strain number and host species have been summarized in Table 1. For genomic DNA extraction, the growing mycelia of individual Hypocreales fungi were purified and cultivated on potato dextrose agar medium (PDA, Oxoid, Basingstoke, UK) for seven days at controlled conditions (20/25 • C, 12 h photoperiod). Harvested mycelia were stored at −20 • C. For qualitative and quantitative cyclodepsipeptide analyses, fourteen day old pure rice cultures of each fungal species were used, according to the procedure described previously [53].

DNA Extraction, Molecular Identification, PCR Primers and DNA Sequencing
Genomic DNA extraction from fungal samples was performed using a modified method with the CTAB (hexadecyltrimethylammonium bromide), according to Gorczyca et al. [61]. The DNA concentration in extracts was quantified using an ND-1000 spectrophotometer (Thermo Scientific, West Palm Beach, FL, USA) and stored in a freezer at −20 • C. For Fusarium species identification, the sequence analysis of a variable fragment of the translation elongation factor 1α gene (tef-1α) was performed, whereas for the other species, the molecular identification was carried out on the basis of the sequence analysis of the internal transcribed spacers of the ribosomal DNA region (ITS1-ITS2).
The In order to analyze the sequences, amplicons were purified with exonuclease I (Thermo Scientific, Espoo, Finland) and FastAP shrimp alkaline phosphatase (Thermo Scientific, Espoo, Finland), afterwards they were labeled using forward primer and the BigDyeTerminator 3.1 kit (Applied Biosystems, Foster City, CA, USA). Finally, PCR-amplified DNA fragments were precipitated with 96% ethanol according to Tomczyk et al. [49].

Sequence Analysis and Phylogeny Reconstruction
Individual DNA sequences were aligned using MUSCLE 3.8.31, available via SeaView 4.6.4 environment [64] (Linux 64bit version). In the case of the BEAS coding sequence, the sequences were first translated and aligned as proteins, then backtranslated into codon alignments. In the case of TEF, exon boundaries were corrected manually to match known splicing site consensus (GT-AG type introns). To investigate grouping of obtained sequences, their top BLAST hits (discontiguous MegaBLAST was performed on the non-redundant GenBank database through NCBI/BLAST stand-alone service, accessed on 22 May 2019) were trimmed appropriately and were also included in the final alignments. In the case of BEAS homologs, additional Beauveria/Cordyceps sensu lato representatives were retrieved from Ensembl/Fungi database (accessed on 17 May 2019).
The phylogeny reconstructions were carried out with IQTREE 1.6.9 [65] (Linux 64-bit version) using automated model fitting, partitioning and ultrafast bootstrap [66]. During the reconstruction, ITS rDNA was modelled as non-coding DNA sequence, while BEAS was investigated using codon-based models. In the case of the translation elongation factor (tef -1α), the coding fragment was investigated as a single coding partition, while the DNA model was ascertained for intronic parts present in the amplified fragments.

Liquid Chromatography Mass Spectrometry Analyses
Culture extracts were screened for the presence of ENNs, BEA, BEAE and ALLOBEA analogues using an LCQ Fleet ion trap mass spectrometer (Thermo Fisher Scientific) coupled to a Waters Acquity UPLC (Milford, MA, USA) with chromatographic column-SunFire C18; instrumental conditions were described earlier by Urbaniak et al. 2019 [3].
Quantitative analysis of mycotoxins (BEA and ENNs) were performed using an LC-MS/MS instrument consisting of an UPLC™ system (Acquity, Waters, Milford, MA, USA) coupled to a triple quadrupole mass spectrometer (TQD, Waters Micromass, Manchester, UK), equipped with an electrospray ionization interface according to Stanciu et al. [67] with our own modifications, described below. The separation of beauvericin and enniatins was achieved using a BEH C18 chromatographic column (100 × 2.1 mm i.d., 1.7 µm particle size; Waters) with injection volume-3 µL and flow rate of mobile phase-0.3 mL/min. The mobile phase consisted of water (A) and methanol (B), both containing 5 mM ammonium formate and 0.1% (v/v) formic acid. The gradient program was as follows: initial conditions at 80% A, 20% B for 2 min; then, from 20% to 90% B in 5 min; next, 90% B for 6 min; from 90% to 100% B in 3 min; return to initial conditions in 2 min. For data processing, Empower TM 1 software was used (Waters, Manchester, UK). Chromatographic parameters are shown in Table 5. Table 5. Parent and daughter ions, collision energy and limit of detection (LOD) and quantification (LOQ) (ng/g) for mycotoxins.

Conclusions
The present study demonstrates the high variability of naturally produced cyclohexadepsipeptides by Hypocreales fungi. It also emphasizes that different fungi have independently evolved a significant level of divergence inside the non-ribosomal peptide synthase genes involved in depsipeptide biosynthesis, along with functional adaptation towards biosynthesis of overlapping yet diversified metabolite profiles. Additionally, a new compound has been described as a naturally occurring beauvenniatin analogue (10).