1. Introduction
The endophytic fungus
Alternaria (section
Undifilum)
oxytropis (Q. Wang, Nagao and Kakish.) Woudenb and Crous produces many secondary metabolites, including swainsonine, a compound toxic to mammals. This and related
Alternaria sect.
Undifilum spp. (family Pleosporaceae, order Pleosporales, class Dothideomycetes) are found within locoweeds,
Astragalus and
Oxytropis species that contain swainsonine. When these plants are ingested by mammals such as cattle, horses, sheep, and goats, normal cellular function is disrupted resulting in locoism disease ([
1,
2]. Other fungi, including
Slafractonia leguminicola (order Pleosporales),
Metarhizium anisopliae (class Sordariomycetes), members of the Arthodermataceae, and a yet undescribed Chaetothyrialean endophyte from
Ipomoea carnea also produce swainsonine [
3,
4,
5]. The biosynthetic pathway for swainsonine has been partially characterized that includes a gene cluster SWN which include a hybrid nonribosomal peptide synthetase (NRPS)-reducing ketide synthase
swnK that was demonstrated to be essential for swainsonine biosynthesis in
Metarhizium robertsii using a knockout of the gene through homologous gene replacement that lacked swainsonine production and complementation that restored its presence [
6]. This work identified a SWN gene cluster associated with swainsonine biosynthesis in
Alternaria oxytropis, S. leguminicola, Arthroderma and the Chaeothyriales fungus associated with
Ipomoea carnea, and showed that the order of the different domains within
swnK was identical among all the swainsonine-producing fungi, including the Chaetothyriales endophyte from
Ipomoea carnea,
Metarhizium spp., and the dermatophytes. Noor et al. [
7] demonstrated that the KS portion of PKS-NRPS is highly conserved among all swainsonine-producing
Alternaria spp., but differed slightly between plant pathogens and nonpathogens.
Many fungi produce specialized secondary metabolites (SM) for virulence, defense, or communication [
8]. Biosynthetic pathways of many SMs include steps catalyzed by multifunctional polyketide synthase (PKS) enzymes. These enzymes direct the structure and biosynthesis of the compounds produced and contain multiple domains leading to a high diversity of SMs [
9]. The high diversity in SMs obscures understanding of how SM pathways have evolved in fungi. In addition to duplications, domain shuffling, neofunctionalization, and subfunctionalization, some SM gene clusters are suggested to be transferred between diverse fungi by horizontal (lateral) gene transfer (HGT) [
8].
Genes associated with toxin biosynthetic pathways in fungi have been found to encode PKSs and NRPSs [
10,
11]. They are often clustered with genes encoding other steps in the same pathway and can be identified from genomes using prediction software based on PKS domains [
12]. Different functional types of PKSs can be identified from sequence and domain structure. Fungi can have type I PKSs, of which there can be highly reducing (HR), non-reducing (NR), or partially reducing (PR) categories, all of which begin with simple carboxylic acids and are iterative [
13]. In addition, fungi have NRPSs that use amino acids as starter units for building compounds. Hybrids involving both PKS and NRPS domains have also been identified in fungi and catalyze production of many different polyketide and amino acid hybrid compounds [
14].
Genome mining of toxin-producing
Fusarium spp. (class Sordariomycetes) for SM genes has revealed 30, 32, 28, and 26 PKS and NRPS gene clusters for
F. graminearum,
F. verticillioides,
F. oxysporum, and
F. solani, respectively [
15]. Examining the genomes of
Aspergillus spp. has also revealed a range of genes. Using multiple software programs along with manual annotations, total clusters (including NRPS-like enzymes) predicted for
A. nidulans,
A. fumigatus,
A. niger, and
A. oryzae are 71, 39, 81, and 75, respectively [
16]. SM genes have also been identified through genome analysis for the endophytic fungus
Pestalotiopsis fici [
17]. In
P. fici, 27 PKSs, 12 NRPSs, and five PKS-NRPS hybrids have been identified [
18]. Perhaps due to their growth habits, endophytes may have smaller numbers of PKSs compared to
Fusarium and
Aspergillus.
To better understand the origins of PKS, NRPS, and PKS-NRPS genes, we investigated their presence in the swainsonine-producing fungus, Alternaria oxytropis. We sought to use genomic analyses to identify PKSs, NRPSs, and hybrid PKS-NRPSs, and predict their structures and functions. The evolutionary relationships among these genes suggest diverse origins for SMs.
2. Materials and Methods
The genome of
A. oxytropis isolated from
Oxytropis sericea collected from Raft River, UT in 1979 was sequenced using Illumina MiSeq with 250 bp paired-end reads on a 400 bp library in Advanced Genetics Technology Center at the University of Kentucky as reported in Cook et al. [
6]. Assembly was done using CLC Genomics Workbench 8.0.2 (Qiagen, Germantown, MD, USA). An unannotated, partially assembled genome (due to high number of repeated sequence) was obtained. For this work, the partially assembled genome was made into a searchable database in Geneious v. R8 (Geneious, CA, USA) [
19]. To identify contigs containing PKS genes, several known nucleotide sequences were used to search the database using the parameters of a discontiguous megablast. Thirty nucleotide sequences of fungal PKS genes were used as query for matching contigs in the
A.
oxytropis genome including three obtained from the biosystems database at the National Center for Biotechnology Information (NCBI) and 27 from Clustermine 360, the database of microbial PKS/NRPS [
20,
21].
Each contig that was identified from the initial search was analyzed for nucleotide sequence similarity using a discontiguous megablast search function of the nucleotide redundant database of NCBI. Contigs containing partial sequence of PKS genes were completed by searching the
A. oxytropis genome for similar matches and manual assembly of contigs within the Geneious software. Maximum parsimony analyses of
A. oxytropis PKS gene contigs along with Clustermine360 fungal PKS genes was used to identify duplicate sequences and determine PKS types (Paup, 1000 bootstrap). All contigs containing
Pks-like genes were characterized through BLASTn, BLASTp, Smartblast, and Conserved Domain Database searches of the NCBI database along with analysis through antibiotics and Secondary Metabolite Analysis SHell (antiSmash) v. 3.0 and Secondary Metabolites Unknown Regions Finder (SMURF) [
22].
Various programs were used to analyze NRPS and PKS genes and domains. AntiSmash was also used to predict NRPS, PKS and secondary metabolites [
23,
24]. Natural Product Domain Seeker (NaPDoS) was used to identify and predict KS and C domains [
25], and Non-ribosomal Peptide Synthase Substrate Predictor (NRPSsp) was used to predict NRPS substrates [
26]. InterProScan was used to form predictive models [
27], and Secondary Metabolite Prediction and Identification (SeMPI) was used to predict and identify pipelines for PKS and NRPS [
28].
Confirmation of sequence for selected sequences was accomplished using PCR and by comparison with a second partial genome sequence of
A. oxytropis. Primers were designed based on the obtained sequences in Geneious, and PCR was performed as previously reported [
29]. All sequences were deposited into Genbank and accession numbers are listed in
Table 1. Closest matches by BLASTp were based on percentage identity for sequences with query coverage greater that 90%.
The highest seven matches for each amino acid sequence using BLASTp were used for generating phylogenetic trees. Sequences were aligned with MUSCLE using Geneious 10.0.9 software and trees were generated using PAUP and maximum parsimony with 100 replicates. Outgroups are marked on each tree and were the sixth closest matches for each tree.
4. Discussion
The number of PKSs identified in this work for
A. oxytropis, 22, is lower compared to some toxin-producing fungi.
Aspergillus species have been identified with high numbers of PKS and NRPS gene clusters (33–81) [
8,
10,
16] and
Fusarium species with 26–32 gene clusters [
15]. Lu et al. [
35] found motifs for 59 possible PKSs for
A. oxytropis, but did not further confirm or identify them, and thus the higher number could be due to pseudogenes.
The presence of KSD and CD (
Table 2) correlated with the presence of PKS, and condensation domains as shown in the figures. The contigs (
Supplemental Figures and
Table 2, show the presence of a KS domain. Contigs from
Figures S2–S4 and S15–S20, and
Table 2 contain a condensation domain. Both C and KS domains are highly conserved. Type I PKS, modular and hybrid, are described by C and KS domain phylogeny [
25].
Cyclization of final peptide product requires the presence of a condensation (C), adenylation (A), and end with a thiolation (T) domains. Fungi NRPS lack a terminal thioester (TE) domain. Instead, macrocyclic fungal NRPS end with a condensation-like domain (Ct), which corresponds, in function, to TE [
36]. Requirements for cyclization include the presence of a C domain, A domain, T domain (PCP, peptidyl carrier protein), and a terminal Ct domain (with or without an attached PP arm) [
36,
37,
38,
39]. Cyclization predictions were confirmed for the NRPS in which a product had been identified. NRPS 33635 satisfied all requirements for cyclization and the product is cyclized. NRPS 8194 is a partial sequence, and its closest matches fulfills the cyclization requirements in producing a cyclic tetrapeptide. Similarly, NRPS 5682 is a partial sequence, and its closest match fulfills the cyclization requirements, however the identity of the product has not been confirmed. Though some other contigs included all three domains required for cyclization, they did not end with a Ct domain, or lacked a PCP domain. NRPS 40703 (
Figure S19), lacks a PP (PCP) domain, but has an ACP-like domain. While ACP domains (acyl-carrier protein) resemble the T domain of NRPS [
40] it is not required for cyclization, which suggests that this contig would not enzymatically cyclize its end product.
The A domain of NRPS are required for the biosynthesis of the peptide natural product, and substrate selection is determined by the A domain. All basic NRPS must have an A domain, required for the biosynthesis of the peptide natural product, and substrate selection; a PCP (thiolation) domain, transports and attaches the substrates to different catalytic domains; and a C domain, that catalyzes peptide bond formation [
34,
40]. While amino acids substrates were predicted, since few final products have been identified, it was difficult to determine the accuracy of some of the predictions.
Seven of the 22 PKSs identified were most similar to an Alternaria species. This was not unexpected as A. oxytropis is most closely related to other Alternaria species and Alternaria oxytropis falls within the Pleosporaceae. Among the seven PKSs most similar to other Alternaria species, they were distributed among the highly reducing PKS, partially reducing PKS, nonreducing PKS, and NRPS. Other fungi with closest matches within the Pleosporaceae were Pyrenophora, Bipolaris, and Setsosphaeria. There were four PKS from the genus Pyrenophora. Bipolaris spp. provided the closest match for two PKS. Fusarium with two matches was the fungus outside of the Pleosporaceae with the highest number of matches.
Several of the PKS/NRPS identified (PKS 2122, PKS 8407, NRPS 5682) showed high identity with Alternaria species or relatives through the first seven matches and three others, PKS 40283, PKS 96133, and NRPS 33635) showed high identity with Alternaria species and other Pleosporales. These PKS and NRPS are obviously highly conserved among Alternaria sp. and Pleosporales. For several of the sequences, including PKS 8407, NRPS 3365, A. oxytropis clustered with other Alternaria spp. in the phylogenetic trees. In other phylogenetic trees (PKS 40283, PKS 96133) A. oxytropis did not cluster with other Alternaria spp.
For
A. oxytropis, 68% of the SM genes did not match the genus
Alternaria even at the amino acid level. Four sequences (PKS-NRPS 58882, PKS-NRPS 21438, NRPS 8194, and PKS 39849) had moderately high identity with
Pyrenophora species in the family Pleosporaceae, and the rest of the matched fungi did not fall within the order Pleosporales. Two of those (PKS-NRPS 21438 and NRPS 8194) showed high conservation with 87–88% aa identity with
Pyrenophora spp. Most
Pyrenophora are pathogenic on cereals and the genus is monophyletic [
41].
Pyrenophora is a sexual state of what was previously known as
Drechslera. Even for sequences in which there is high identity with
Alternaria species, such as PKS 2122 and NRPS 5682, the phylogenetic trees suggest that
A. oxytropis is as closely related to
Pyrenophora as to
Alternaria spp. This raises questions about the inheritance of these SM and the taxonomic placement of the fungus within the genus
Alternaria. This high sequence conservation suggests a potential taxonomic lineage for
Alternaria oxytropis, perhaps as a product of hybridization or recombination between
Alternaria and
Pyrenophora ancestors.
Three sequences analyzed, PKS 9132, PKS 17612, and PKS 3398, had no closest matches in the Pleosporales. For the PKS 9132 sequence, the phylogenetic trees show no clustering of A. oxytropis with any of the closest matches, whereas for the other two sequences, A. oxytropis clustered with otherwise unrelated fungi.
Within the phylum Ascomycota, the large majority of SM gene clusters are found only in closely related species [
8]. Those few that are more broadly distributed across the phylum are often highly divergent between even closely related species [
22]. Rokas et al. [
42] suggested several possible explanations for SM cluster variability based on molecular evolutionary processes including functional diversity, horizontal gene transfer, and de novo assembly. For
A. oxytropis SM, functional diversity is likely, horizontal gene transfer is possible, and no evidence was found for de novo assembly.
Orthologous or paralagous functional diversity could explain some of the variability found in
A. oxytropis SM, particularly for those with the closest matches to members of the Pleosporaceae, but not
Alternaria, marked as P in
Table 1. All five of these SM genes showed slight divergence in domain size or order from their closest matches. NRPS 8194 differs slightly in gene order from that of the HC-toxin synthetase gene of
Pyrenophora tritici-repentis and
A. oxytropis is not known to produce HC toxin. The SWN gene cluster of swainsonine-producing fungi vary in the order of genes other than
swnK (PKS-NRPS 58882) in the cluster [
6].
The
A. oxytropis SM genes that did not match fungi within the order Pleosporales or did not match fungi within the class Dotheomycetes have a more speculative ontogeny. Some of the variability could be to due to horizontal gene transfer or more likely a combination of horizontal gene transfer and orthologous or paralagous changes. The domains (and order of domains) in these
A. oxytropis SM genes are similar to their closest matching fungi. Horizontal gene transfer between fungal classes has been speculated for the entire sterigmatocystin biosynthetic pathway, that appears to have been transferred from
Aspergillus nidulans to
Podospora [
43]. A study of SM within
Aspergillus fumigatus showed that 13 of the SM gene clusters were generally conserved with low variation and 23 were highly variable [
8]. Those authors found six examples of gene content polymorphisms that were exemplified by loss of gene cluster function, structural changes in the metabolite, or change in the expression or transport of the metabolite.
Manning et al. [
44] reported that the genome of
Pyrenophora tritici-repentis includes several NRPSs that may have been derived by horizontal gene transfer and gene duplication. The polyketide host-selective toxins associated with
Alternaria spp. adapted to pear, apple, tangerine, citrus, rough lemon, and tomato are all found on conditionally dispensable chromosomes (CDC) that could have been acquired through horizontal chromosome transfer among the
Alternaria spp. [
45]. Armitage et al. [
46] showed that while
A. tenuissima pathotypes shared 10 types of transposable elements (TE) with
A. arborescens, the pathogens contained significantly decreased numbers of TE in the DDE and gypsy families, and significantly higher numbers of TE in the mariner family. They speculated that the TEs may have increased the variability in the fungi. Determining if gain of whole gene clusters through TE or horizontal chromosome transfer might be possible explanations for SM sequences in
A. oxytropis would require many more sequences from related species, chromosome mapping of the fungus, analysis of the number and location of TE, and functional analyses of the SM.
The closest matches for the
A. oxytropis PKSs were for
Alternaria species for highly conserved genes. PKS40283 had the highest amino acid identity (95%) of any of the PKSs identified, with
A. alternata PKSA, which is responsible for melanin production. The gene responsible for melanin production is highly conserved among many dark-colored fungi. Although
A. oxytropis is dark black in culture and
A. cinerea and
A. fulva are tan and grey, respectively, in culture [
29], the latter two fungi have the same sequence as identified for PKS40283. Silencing of the homologous gene,
pks1, in
Slafractonia leguminicola, also within the family Plesoporaceae and a swainsonine toxin producer, caused a reduction in melanin synthesis and relevant transcript levels [
47].
NRPS 3365 also had high amino acid homology (91%) with
Alternaria alternata NPS6. This is likely a highly conserved NRPS among both pathogenic and saprobic fungi, including pathogenic and nonpathogenic isolates of
Pyrenophora tritici-repentis [
44]. NPS6 of
Cochliobolus spp. was found to have orthologs in all filamentous fungi examined [
48]. It was also found to affect virulence since loss of expression reduced virulence but did not completely abolish it. Additionally, lack of expression led to an increase in sensitivity to hydrogen peroxide. In
Alternaria alternata,
nps6 is necessary for the biosynthesis of dimethyl copreogen siderophores as well as functioning as a virulence factor [
49]. This was the predicted SM for
A. oxytropis for NRPS 3365 as well. Multiple transcription factors control
nps6 transcript accumulation in
A. alternata including NADPH oxidase, a redox responsive transcription facto YAP1 and a mitogen activated protein kinase HOG1 [
50]. Mutation of
nps6 or any of the transcription factors results in increased sensitivity to reactive oxygen species (ROS) and reduced virulence in citrus. ROS sensitivity can be partially rescued though with the addition of iron. The effects of iron and ROS sensitivity lead to the idea that NPS6 is important to the production of siderophores involved in iron uptake [
50]. Deletion of NPS6 in
Fusarium graminearum,
Cochliobolus miyabeanus, and
A. brassicicola also resulted in the same responses of increased sensitivity to ROS and reduced virulence [
51].
NRPS 5682 also showed the same high level of amino acid identity (91%) with
Alternaria alternata NRPS1.
Alternaria NRPS1 is likely involved in plant infection since increased expression of
nrps1 was found during host infection by
A. brassicae [
52]. This NRPS could play some role in the establishment of the endophytic relationship of
A. oxytropis in its plant host. PKS2122 showed a high level of amino acid identity (90%) with
Alternaria solani PKSN. PKSN is the product of the
alt5 gene, which is essential for alterapyrone biosynthesis [
30]; when
alt5 was expressed in
Aspergillus oryzae under an alpha-amylase promoter, alternapyrone was produced. Because there was a high level of identity at both the nucleotide and protein levels between PKS2122 and PKSN, it is likely that
A.
oxytropis also produces this compound. PKS2122 was very highly conserved among many
Alternaria species. PKS-NRPS 5882 was verified as the
swnK KS [
6]. Two of the fungi with high amino acid identity,
Pyrenophora seminiperda and
Clohesyomyces aquaticus have not been reported to produce swainsonine or be associated with mammalian toxicity.
Both the number of PKS, along with the lack of associated functions identified here, are common to studies in other fungi. While additional PKS genes or PKS-like genes might be identified from the completed genome sequence of A. oxytropis, it is likely that the majority of the PKS genes have been identified here. Interestingly, the functions of the 22 genes found in this study remain unknown. Although secondary metabolites have been studied in many fungi, and the genomes of many fungi have been sequenced, it is significant that functions for these have not been readily identified. Transcriptomic and metabolomic studies will likely be necessary to better understand the function of the genes identified in this work.