Fumonisin and Beauvericin Chemotypes and Genotypes of the Sister Species Fusarium subglutinans and Fusarium temperatum

Fusarium subglutinans and F. temperatum are sister species and maize pathogens commonly isolated worldwide that can produce several mycotoxins and cause seedling disease, stalk rot, and ear rot. The ability of these species to produce beauvericin and fumonisin mycotoxins is not settled, as reports of toxin production are not concordant at the species level. Our results are consistent with previous reports that strains of F. subglutinans produce neither fumonisins nor beauvericin. The status of toxin production by F. temperatum needs further work. Our strains of F. temperatum did not produce fumonisins, while some strains produced beauvericin and others did not. These results enable more accurate risk assessments of potential mycotoxin contamination if strains of these species are present. The nature of the genetic inactivation of BEA1 is consistent with its relatively recent occurrence and the close phylogenetic relationship of the two sister species.

KEYWORDS comparative genomics, gene inactivation, maize pathogens, mycotoxin biosynthesis, secondary metabolism, Argentina, Fusarium subglutinans, Fusarium temperatum, UPLC-MS, beauvericin, fumonisins, genome sequence, mycotoxins, plant pathogens F usarium subglutinans is an important pathogen of maize commonly isolated worldwide and is considered a causal agent of seedling disease, stalk rot, and ear rot (1). This species also can produce a broad range of mycotoxins (2). Within the morphological F. subglutinans sensu lato species, two populations were identified based on DNA sequence data (3). The two populations, F. subglutinans group 1 and F. subglutinans group 2, appeared to be reproductively isolated in nature and were presumed to be in the process of sympatric genetic divergence (3). Fusarium subglutinans group 1 has now been formally described as Fusarium temperatum (4), while F. subglutinans group 2 has retained the formal Fusarium subglutinans sensu stricto name.
Mycotoxin production by these species is of particular interest because production of beauvericin, a cyclic hexadepsipeptide with insecticidal and carcinogenic properties (5)(6)(7), has been reliably reported only in F. temperatum (group 1) and not in F. subglutinans (8)(9)(10)(11). Beauvericin production has been used to identify the species to which some strains belong (11). Continuing studies of F. temperatum and F. subglutinans on cereals, primarily maize (12)(13)(14)(15)(16)(17)(18)(19)(20), have resulted in a general consensus that beauvericin is produced only by strains of F. temperatum and not by strains of F. subglutinans, but the genetics underlying these differences has not been investigated in any detail. Differences in beauvericin production by these two closely related species could provide insights into the evolutionary processes involved in their separation into different species.
The beauvericin (Bea) biosynthetic gene cluster was first described in Fusarium fujikuroi IMI 58289 and consists of a four-gene cluster: Bea1, which encodes the NRPS22, the nonribosomal peptide synthase responsible for synthesizing the beauvericin backbone, and Bea2, Bea3, and Bea4, which encode proteins with transport and regulatory functions (21). Orthologous four-gene biosynthetic clusters also are known in Fusarium proliferatum, Fusarium mangiferae, and Fusarium oxysporum (21), all of which are reported as beauvericin producers in multiple studies (22,23). F. proliferatum is a common contaminant of cereals such as maize, wheat, and barley and can contaminate these substrates with beauvericin as well (22). F. mangiferae is a major cause of mango malformation worldwide (24), but a role for beauvericin in its phytotoxicity has not yet been identified. In F. oxysporum, a causal agent of tomato wilt, beauvericin reduces the level of ascorbic acid in the tomato cells, leading to the collapse of the ascorbate system and protoplast death (25).
The fumonisin (Fum) biosynthetic gene cluster in the genus Fusarium has been well described and includes 16 genes that encode biosynthetic enzymes and regulatory and transport proteins. Functions of genes in fumonisin biosynthesis have been determined in Fusarium verticillioides (26), and the number, order, and genomic orientation of the Fum genes are known in F. proliferatum and F. oxysporum (27)(28)(29). Sequences flanking the Fum gene cluster differ among species, however, indicating that the cluster's genomic location is species dependent (26). Reports of fumonisin production on cracked corn (10,14,20,30,31) by some strains of F. temperatum and F. subglutinans are inconsistent with reported genetic capabilities for fumonisin biosynthesis by these species as sequenced strains of both F. subglutinans and F. temperatum lack one or more of the Fum genes required for fumonisin biosynthesis (26,28,32).
The objectives of this study were to further test the ability of these species to synthesize beauvericin and/or fumonisin with definitive chemical tests of strains not cultured on cracked corn and genetic analyses of additional strains. Our working hypotheses were the following: (i) that no strains of either species could synthesize fumonisin, (ii) that F. temperatum strains, but not those of F. subglutinans, could synthesize beauvericin, and (iii) that the chemical phenotypes would be consistent with the genomic sequence genotypes. The study advances the field by providing new insights into the toxigenic potential of these species and enabling more accurate estimation of the risks they pose to the food and feed products they might contaminate. (

RESULTS
Strain isolation and identification. Twenty-five Fusarium strains from Argentina (Table 1) were identified to species level in a maximum likelihood (ML) phylogenetic analysis of a three-gene combined data set, including sequences of reference strains from related species, with F. proliferatum NRRL 62905 as the outgroup (Fig. 1). Twelve strains were contained within a well-supported clade (bootstrap value, 88) that included the F. temperatum reference strain ITEM 16196 (MUCL 52463) (4). The remaining 13 strains were contained within a second well-defined clade (bootstrap value, 99) that included the F. subglutinans reference strain NRRL 22016 (Fig. 1).
Genome analyses. (i) Genome assemblies. We generated genome assemblies for two strains of F. subglutinans (RC 298 and RC 528) and one strain of F. temperatum (RC 1°C) mean annual temperature and 500 to 800 mm of annual precipitation, respectively. NOA1 contains four locations across Quebrada de Humahuaca in the Jujuy province, with an 11.7°C (5.1 to 16.3°C) mean annual temperature and 400 mm of annual precipitation. c Range, 7 to 400 g/kg; mean production, 71 g/kg; median production, 11 g/kg. d Information on ITEM strains is available on line at http://www.ispa.cnr.it/Collection/. 2914) ( Table 2). For F. temperatum RC 2914, ϳ7.7 million reads were assembled in 720 scaffolds, for a total length of 42.5 Mb when only scaffolds of Ն10 kb in length were included. The scaffold N 50 , i.e., the length of the shortest scaffold such that 50% of the assembly is found in scaffolds of this length or longer, was 334 kb, and the longest scaffold was 1.5 Mb. The average coverage was 53ϫ. Two scaffolds were retained for analysis of the Bea and Fum clusters.
For F. subglutinans RC 298, ϳ17 million reads were assembled in 4,088 scaffolds, for a total length of 49.7 Mb when only scaffolds of Ն10 kb in length were included. The scaffold N 50 was 228 kb, and the largest scaffold was 975 kb. The average coverage was 101ϫ. Two scaffolds were retained for analysis of the Bea and Fum clusters. Finally, for F. subglutinans RC 528, ϳ9 million reads were assembled in 1,418 scaffolds, for a total length of 43 Mb when only scaffolds of Ն10 kb in length were included. The scaffold N 50 was 204 kb, and the largest scaffold was 997 kb. The average coverage was 61ϫ. Again, two scaffolds were retained for analysis of the Bea and Fum clusters.
(ii) Genomic context of contigs containing the beauvericin and fumonisin clusters. Dot plot analysis between chromosome 9 of F. fujikuroi IMI 58289 (Ffuj_Chr9),  where a complete Bea cluster is located, and scaffold 7 of F. temperatum CMWF 389 (Ftemp_Scaff7) identified sequences of almost the same length with complete synteny. Thus, Ftemp_Scaff7 probably is orthologous to chromosome 9 predicted for F. fujikuroi (Fig. 2). Dot plot analysis between chromosome 1 of F. verticillioides FGSC 7600 (Fv_Chr1), where the Fum cluster is located, and two scaffolds of F. temperatum CMWF 389, scaffold 1 (Ftemp_Scaff1) and scaffold 12 (Ftemp_Scaff12), had very good synteny. Thus, chromosome 1 of F. verticillioides is orthologous to Ftemp_Scaff1 and Ftemp_Scaff12 (Fig. 3). Circos plot analysis with the complete Bea cluster from Ffuj_Chr9 and portions of Ftemp_Scaff7 and the three newly sequenced Bea-containing contigs shows that the Bea cluster is complete in both F. temperatum and F. subglutinans (Fig. 4A).
Circos plot analysis with the complete Fum cluster from F. verticillioides chromosome 1 and portions of contigs from the three newly sequenced strains shows a gap in the synteny. Thus, both F. temperatum and F. subglutinans lack most of the genes normally found in this biosynthetic cluster (Fig. 4B).
Beauvericin cluster. The entire Bea cluster (21) is present in the F. subglutinans and F. temperatum strains sequenced in the current study, as well as in several other closely  related species that produce beauvericin, e.g., F. fujikuroi, F. mangiferae, Fusarium nygamai, F. oxysporum, and F. proliferatum (Fig. 5). In Fusarium circinatum FSP 34, the Zn(II) 2 Cys 6 transcription factor (FFUJ_09298), encoded by Bea4, is absent, and this gene also is missing in the other two F. circinatum genomes in GenBank (strains GL 1327 and KS 17).
Complete and functional BEA2, BEA3, and BEA4 proteins are predicted for all three genomes assembled in this study. The Bea1 gene encoding the nonribosomal peptide synthase NRPS22 is predicted to produce a functional protein in both F. temperatum strains (RC 2914 and CMWF 389). In F. subglutinans, the predicted protein is apparently nonfunctional in strain RC 528 due to a single nucleotide polymorphism (SNP) resulting in a premature stop codon (CAG ¡ TAG transition; the SNP is underlined). This transition occurs at nucleotide position 7685 (relative to the Bea1 sequence from F. fujikuroi IMI 58289), where position 1 coincides with the start of the reading frame, i.e., the adenine of the ATG start codon. In strain RC 298, there is an insertion of a single cytosine at position 5875 that results in a frameshift and premature truncation of the protein (Fig. 6).
Both of the F. subglutinans strains had a 184-bp insertion between nucleotides 4223 and 4416 (Fig. 6). If this insertion was transcribed, it would add 61 amino acids to the length of the protein and cause a frameshift in the downstream reading frame that would lead to premature truncation of the protein. In silico prediction programs exclude the 184-bp insertion region from the open reading frame and instead introduce novel introns to prevent the premature truncation of the protein due to in-frame stop codons within the insertion. This predicted gene transcript would still result in a large protein, but it is uncertain whether the resulting protein would function properly.
Fumonisin cluster. The entire Fum cluster was missing from the F. subglutinans and F. temperatum genomes, which is consistent with the reported inability of many strains of these species to produce fumonisins. We searched for portions of all 16 Fum cluster genes (26) but found no recognizable homologous sequences. Fusarium subglutinans and F. temperatum are members of the American clade of the F. fujikuroi species complex (FFSC). Some members of this clade, e.g., Fusarium anthophilum and Fusarium bulbicola, can produce fumonisins and carry the Fum biosynthetic gene cluster (26,32,33). We queried our newly generated genomes and those of some other members of the FFSC with genes that flank the Fum cluster in species from all three clades of the FFSC (26). In all cases, the Fum cluster was absent from F. subglutinans and F. temperatum. Instead, we found one of four flanking genes (Cpm2) from the American clade species and two of four genes from Asian clade species (Mfs1 and Zcb1). We also found all four genes queried from African clade species (Znf1, Zbd1, Orf20, and Orf21) although the orientations and order of Orf21 and Znf1 were different in F. subglutinans and F. temperatum from those in F. verticillioides (Fig. 7).  Mycotoxin production. Nine of the 12 strains identified as F. temperatum produced beauvericin at levels ranging from 7 to 400 g/kg (mean, 71 g/kg; median, 11.3 g/ kg), whereas no F. subglutinans strains produced beauvericin. None of the 25 strains examined produced fumonisin B 1 (FB 1 ) on potato dextrose agar (PDA) ( Table 1).

DISCUSSION
Fusarium subglutinans and F. temperatum are well known as preharvest fungal pathogens that cause maize stalk and ear rot and are closely related species that can be easily misidentified (4). Strains of these species can produce a variety of mycotoxins (2,8,10,14,30,31,34). However, reports of mycotoxin production by these species are not consistent (14,20), leading to confusion regarding the specific mycotoxin profile that they possess. This confusion can result in underestimation or overestimation of the mycotoxin-associated risk posed by foods and feeds contaminated with these fungi. It also makes it very difficult to develop effective pre-and postharvest strategies for monitoring and managing mycotoxin contamination.
There are multiple reports of fumonisin production (10, 14, 30, 31) and nonproduction (2, 14, 34, 35) by F. subglutinans groups 1 and 2, which are now F. temperatum and F. subglutinans, respectively. The lack of all or parts of the Fum gene cluster in some strains of both species has been reported on multiple occasions (26,27,32). In our study, we found that some genomes of both species lacked the entire Fum cluster and that the insertion sites across species in the FFSC that contain part or all of the Fum gene cluster are not well conserved. For example, in Fusarium musae, a sister species of F. verticillioides that cannot produce fumonisins (36,37), only remnants of the Fum21 and Fum19 genes, at the opposite ends of the cluster, remain along with some of the flanking genes. The deletions and rearrangements we detected in genomic regions where the Fum cluster is inserted in other species suggest that changes related to Fum cluster insertion/deletion are not simple events and could have occurred in more than one step at more than one time.
In contrast with fumonisins, there is a general consensus that strains of F. subglutinans do not produce beauvericin but that some strains of F. temperatum do (10-12, 14-16, 18). In the present study, we found that 75% of the F. temperatum strains analyzed could produce beauvericin but that none of the strains of F. subglutinans could. Unlike the Fum cluster, however, the molecular basis for the differences between toxin-producing and toxin-nonproducing strains was not previously known.
The Bea gene cluster contains four genes, Bea1 to Bea4, of which two, Bea1 and Bea2, are essential for beauvericin production, while the other two, Bea3 and Bea4, encode proteins that repress beauvericin production (21). BEA4 is not essential for beauvericin production since F. circinatum can synthesize beauvericin (5, 38-40) but lacks the gene encoding this protein (21). In fact, deletion of Bea4 could potentially increase beauvericin production by removing a layer of repressive regulation.
The Bea1-encoded nonribosomal polypeptide (NRPP) synthetase required for biosynthesis of the cyclic depsipeptide beauvericin was first described in the fungus Beauveria bassiana over 50 years ago (41,42) and later confirmed in F. circinatum, F. oxysporum, F. proliferatum, and F. fujikuroi (21,23,39,43,44). Molecular organization of the Bea gene cluster has not been analyzed as extensively as has the Fum gene cluster. The genomic organization of the Bea clusters in F. subglutinans, F. temperatum, F. circinatum, F. proliferatum, F. fujikuroi, F. mangiferae, and F. nygamai is consistent with respect to gene order, direction of transcription, and genomic context; however, there are differences in individual gene coding sequences.
The available F. temperatum genomes are all from beauvericin-producing strains and harbor intact, functional sequences for all of the Bea genes in the cluster. All F. subglutinans genomes carry functional Bea2 to Bea4 genes. The Bea1 gene appears to encode a nonfunctional protein in both of the analyzed sequences from F. subglutinans. Both of these genomes contain a 184-bp insertion at position 4233. This insertion results in a protein projected to be nonfunctional, whether it alters splicing and intron arrangement or is read as a coding part of the gene. Each strain carries a second, but different, mutation that also inactivates the protein. In RC 298, there is a single nucleotide insertion at position 5686 that introduces a frameshift resulting in a stop codon 120 bp further downstream (position 5806) that should prevent translation of a full-length protein. In RC 528, a single nucleotide substitution at position 7685 results in a premature stop codon 1,907 bp upstream of the 3= end of the coding region.
The accumulation of only a couple of loss-of-function mutations in Bea1 suggests that the process of its inactivation began relatively recently. As both strains have the 184-bp insertion, this genomic change probably occurred first. Assuming that this insertion prevents beauvericin accumulation, then subsequent mutations in genes required exclusively for beauvericin biosynthesis would occur without selection acting against them. Thus, the longer a gene has been nonfunctional, the more mutations it should have accumulated in its coding sequence. After the insertion occurred, flawed transcripts might still produce altered proteins. If so, secondary mutations, such as the Bea1 single nucleotide insertion or substitution we observed, could have been selected for to reduce the production of proteins with toxic effects or to reduce the energetic costs due to transcription and translation of nonfunctional genes, speeding the rate at which mutations accumulate (45)(46)(47). Given the difference in secondary mutations seen in the strains sequenced, other strains that do not produce beauvericins could well have other mutations in Bea1 or elsewhere that prevent beauvericin biosynthesis. Yet the few loss-of-function mutants found in either of the two sequenced strains support a recent Bea1 inactivation.
Analysis of transcripts from the mutated gene could provide insights into how F. subglutinans has managed the 184-bp insertion in this gene. For example, are the novel introns predicted in the in silico analysis present? Or is the entire insertion translated, which would result in a single-base frameshift mutation? The F. temperatum strains that do not produce beauvericin could be of interest as well. Do they carry the 184-bp insertion and either of the other mutations observed in the F. subglutinans genomes? Or is their inability to produce beauvericin due to mutations elsewhere in the Bea cluster or the strains' genomes?
The nature of the genomic changes that disrupt mycotoxin production plays a role in the potential development of diagnostic PCR tests for whether strains could potentially produce fumonisins or beauvericin. Strains of both species would be negative if any primer pairs designed to amplify any portion of the Fum cluster were used as the entire cluster is missing from the available genomes. A similar test for the potential to produce beauvericin is more problematic. Both species have all of the genes in the Bea cluster, and the genes Bea2 to Bea4 are predicted to be intact and functional. Thus, any successful DNA-based assay would need to be specific to Bea1. To detect the aberrant F. subglutinans versions of these genes, the assay could have primers that result in a larger fragment due to the 184-bp insertion or have one primer based on a unique sequence within the inserted region. Tests that detected a secondary SNP or the presence of the insertion also could identify nonfunctional alleles. Other PCR tests involving Bea1, i.e., simply detecting the presence of the gene or a portion of it, would be unable to distinguish a functional version of the gene from the nonfunctional version seen in F. subglutinans. Depending on the reason for the inability of the three F. temperatum strains to produce beauvericin, this assay could become even more complex.
In conclusion, we found that 25 strains of F. subglutinans and F. temperatum from Argentina could not synthesize fumonisins. The genomic basis for the lack of fumonisin production is presumably the complete absence of the genes in the Fum cluster, given the available genome sequences. As some F. temperatum strains are reported to produce fumonisins (14,20), however, sequences of genomes from these strains are needed to understand the complexities of mycotoxin production in this species. We also confirmed that all tested strains of F. subglutinans and a subset of F. temperatum strains cannot synthesize beauvericin and note that the lack of beauvericin production cannot be used to definitively identify a strain as F. subglutinans. The Bea cluster was organized consistently in terms of location, gene order, and direction of transcription in F. circinatum, F. fujikuroi, F. subglutinans, and F. temperatum. Potential similarities in the Bea1 sequences from strains of F. subglutinans and the non-toxin-producing strains of F. temperatum could show whether the initial inactivation event preceded the separation of F. subglutinans and F. temperatum as separate species. Since the NRPP responsible for enniatin synthesis differs in only a few amino acids from the NRPP responsible for beauvericin synthesis (48), it will be interesting to determine if events that prevent enniatin synthesis are similar to those that prevent beauvericin synthesis. Our study provides a firm genetic and physiological base on which future studies of these toxins can be built.

MATERIALS AND METHODS
Fungal isolates. Strains of Fusarium were recovered from maize harvested in four regions of Argentina where the presence of F. subglutinans and F. temperatum had previously been reported (14,49). Maize grains were incubated on pentachloronitrobenzene (PCNB) medium (24), and the resulting Fusarium colonies were purified by subculturing single microconidia from them. Morphological identifications were made following growth on homemade (24) and commercial (Biolife, Milan, Italy) potato dextrose agar (PDA), carnation leaf agar (CLA) (24), and Spezieller Nährstoffarmer agar (SNA) (24) for 10 days at 25°C under 12-h alternating periods of light (combination of cool white and black lights) and darkness. Colony morphology was evaluated on PDA. Spore morphology was evaluated using spores Fumero et al. Applied and Environmental Microbiology from colonies growing on CLA or SNA. Strains with the morphological characteristics of F. subglutinans described by Leslie and Summerell (24) were selected for DNA-based identification and further study.

DNA-based identification of fungal isolates. (i) DNA extraction.
Twenty-five strains with morphology consistent with that of F. subglutinans were selected for DNA-based identification. Isolates were grown on PDA for 2 days at 25°C in the dark. Fresh mycelia were collected by scraping the plate surface and collecting the mycelia in 2-ml tubes. Total genomic DNA was extracted from 30 mg of freeze-dried and ground mycelia by using a Wizard Magnetic DNA Purification System for Food kit (Promega, Madison, WI) according to the manufacturer's protocol. DNA was quantified in a NanoDrop spectrophotometer, and the DNA concentration was adjusted to 20 ng/l for PCR amplifications.
(ii) Gene sequencing. Portions of three housekeeping genes, encoding ␤-tubulin (Tub2), translation elongation factor (Tef1), and the second largest subunit of RNA polymerase II (Rpb2), were used for species identification. Previously described PCR conditions and primers were used for each gene: BT2a/BT2b for Tub2 (50), EF1/EF2 for Tef1 (51), and 5F/7cR for Rpb2 (52). PCR amplicons were cleaned before sequencing with EXO/FastAp (exonuclease I, Escherichia coli/FastAP thermosensitive alkaline phosphatase; ThermoFisher Scientific Baltics, Vilnius, Lithuania) to hydrolyze excess primers and nucleotides. Both strands were sequenced with a BigDye Terminator, version 3.1, cycle sequencing ready reaction kit. Sequence reaction products were purified by gel filtration through Sephadex G-50 (5%) (Amersham Pharmacia Biotech, Piscataway, NJ) and analyzed on a 3730xl DNA analyzer (Applied Biosystems, Foster City, CA). The software package Bionumerics, version 5.1 (Applied Maths, Sint-Martens-Latem, Belgium), was used to align the two DNA strands and edit the sequence. Edited sequences were compared with sequences in the Fusarium-ID (53) and GenBank databases. The phylogenetic species identity of each field strain was assigned to the species of database strains when sequence identity was Ͼ98%. NCBI accession numbers for Tef1, Tub2, and Rpb2 sequences for each strain are listed in Table 1.
(iii) Phylogenetic analyses. DNA sequences consisting of partial sequences of Tub2, Tef1, and Rpb2 were concatenated and then aligned with ClustalW. The resulting combined data set was analyzed with the maximum likelihood algorithm implemented in IQ-TREE (54) with the Tamura-Nei substitution model (55) and 1,000 bootstrap replicates (56). The alignment was deposited in TreeBASE (https://www.treebase .org/treebase-web/search/studySearch.html) under study number 25708.
Gene cluster analysis. (i) DNA extraction for whole-genome sequencing. The genomes of two F. subglutinans strains (RC 298 and RC 528) and one F. temperatum strain (RC 2914) were sequenced. Each strain was cultivated in 50 ml of complete medium and incubated on an orbital shaker at 150 rpm for 2 days at 25°C (24). Mycelia were collected following vacuum filtration through nongauze milk filter disks (KenAG, Ashland, OH) and stored at -20°C in 2-ml tubes. Frozen mycelia were lyophilized (Labconco Corporation, Kansas City, MO), added to microcentrifuge tubes containing two 4.5-mm zinc-plated steel beads (Daisy BBs, Rogers, AR), and ground to a fine powder in a mixer mill (Verder Scientific, Retsch, Germany). Genomic DNA was isolated by following a modified cetyltrimethylammonium bromide (CTAB) protocol (24). The resulting DNA was resuspended in Tris-EDTA (TE) buffer (pH 8.0) and stored at -20°C. DNA quality was checked by separation in a 1% agarose gel. DNA concentration was measured with a Quant-iT PicoGreen double-stranded DNA (dsDNA) assay kit (Life Technologies, Carlsbad, CA), and the results were read in a Synergy H1 hybrid reader (BioTek Instruments, Inc., Winooski, VT). The DNA was diluted to a final concentration of 100 ng/l.
(ii) Genome sequencing and assembly. Three paired-end libraries (one for each selected strain) were constructed and sequenced with an Illumina MiSeq sequencer using paired-end 300-bp reads at the Kansas State University Integrated Genomics Facility. Genomes were assembled into contigs by using the de Bruijn graph-based algorithm implemented in the DISCOVAR de novo software from the Broad Institute, Cambridge, MA (https://software.broadinstitute.org/software/discovar/blog/) with the default parameters (k-mer of 200). Fastq files were converted to BAM files with the tools in Picard, version 2.12.1 (http://broadinstitute.github.io/picard). Though the DISCOVAR de novo assembly does not contain long-range scaffolding information, the sequences represented by these fastq files are technically scaffolds due to the presence of some stretches of Ns that bridge small gaps in read coverage. We refer to them as scaffolds although they are functionally more similar to contigs from other assemblies.
The same blastN analysis protocol was used for the Fum gene cluster but with the predicted F. verticillioides Fum gene cluster serving as the reference (FVEG_00316 to FVEG_00329) (27,62). For the Fum cluster, the analysis was extended to regions flanking the cluster by including the genes described by Proctor et al. (26).
Annotation of the Bea biosynthetic genes and Fum flanking genes present in the newly sequenced genomes of F. subglutinans and F. temperatum was done manually, with the gene prediction tools Augustus (63) and FGENESH (64). The locations of coding sequences and introns were determined by comparison with the publicly available annotated sequences of the reference strains.
(iv) Genomic context of newly sequenced contigs containing clusters of interest. The genomic contexts of the putative Bea and Fum clusters in the newly sequenced genomes of F. temperatum RC 2914 and F. subglutinans RC 298 and RC 528 were established. The F. temperatum CMWF 389 (57) genome assembly used as a reference is in the scaffold stage, so dot plots were used to compare these scaffolds with the well-annotated chromosomes of F. verticillioides FGSC 7600 (61) and F. fujikuroi IMI 58289 (59). The online tool Circoletto (http://tools.bat.infspire.org/circoletto/) was run with default parameters (65). The resulting circular plots provide a global view of the sequence similarity between the Bea and Fum gene clusters and flanking regions from reference genomes and the newly sequenced contigs of F. subglutinans and F. temperatum. This software also was used to verify that contigs with blastn hits contained complete sequences of the clusters of interest, or the flanking regions, and to display aspects of the alignments, such as sequence rearrangements and percent identity.
Mycotoxin analysis. (i) Beauvericin and fumonisin B 1 (FB 1 ) production in vitro. Mycotoxins were produced on PDA, as previously described for Fusarium (66). Plates were centrally inoculated with 3-mm-diameter mycelial plugs from the edges of 7-day-old SNA cultures. Inoculated plates were incubated for 15 days in darkness at 25°C. Each plate was inoculated in duplicate. This experiment was performed once.
(ii) Chemicals and preparation of standards. All solvents (high-performance liquid chromatography [HPLC] grade) were purchased from VWR International SRL (Milan, Italy). Ultrapure water was produced by a Millipore Milli-Q system (Millipore, Bedford, MA). Beauvericin standards (purity of Ͼ99%) were purchased from Sigma-Aldrich (Milan, Italy), and FB 1 was from Biopure (Romer Labs Diagnostic GmbH, Getzersdorf, Austria). Standard stock solutions (1 mg/ml) were prepared by dissolving the solid commercial toxin standards in methanol. For working solutions of beauvericin, some of the methanol stock solution was dried under a nitrogen stream at 50°C and reconstituted with methanol-water (70:30, vol/vol). Standard solutions for ultraperformance liquid chromatography (UPLC) calibration were prepared by using different concentrations in a range of 0.02 to 40.00 g/ml. Working stock solutions of FB 1 were prepared by drying some of the stock solution under a nitrogen stream and reconstituting it with acetonitrile-water (1:1, vol/vol). Standard solutions for UPLC calibration were prepared by using different concentrations in a range of 0.01 to 1.00 g/ml. Standard solutions were stored at -20°C and warmed to room temperature (ϳ20 to 22°C) prior to use.
(iii) Determination and confirmation of beauvericin production. Ten grams of culture material was extracted with 15 ml of methanol on an orbital shaker (150 rpm) for 30 min. Six milliliters of the extract was evaporated to dryness under a stream of nitrogen at 40°C. The residue was dissolved in 1.5 ml of methanol-water (70:30, vol/vol) and filtered through a 0.2-m-pore-size regenerated cellulose (RC) filter (Grace Davison Discovery Science, Columbia, MD). Ten microliters of the extract was injected into the full-loop injection system of an Acquity UPLC system (Waters, Milford, MA), equipped with an electrospray ionization (ESI) interface with a binary solvent manager, a sample manager, a column heater, a photodiode array, and quadrupole dalton (QDa) detectors. The analytical column was an Acquity UPLC BEH C 18 (2.10 by 100 mm; 1.7-m particle size) preceded by an Acquity UPLC in-line filter (0.20-m pore size). The temperature of the column was set at 50°C. The flow rate of the mobile phase was set at 0.35 ml/min. The toxins were determined in both detectors, with the photodiode array set at 205 nm, and QDa mass detector (UPLC-PDA-QDa), without splitting. The mobile phase consisted of a gradient with two components: solvent A consisted of water with 0.1% formic acid, and solvent B consisted of acetonitrile with 0.1% formic acid. The initial composition 50:50 (A/B) was kept constant for 2 min; solvent B was then increased linearly to 75% in 8 min, followed by another linear increase to 80% in 2 min, and the composition was kept constant for 4 min. For column reequilibration, solvent B was linearly decreased to 50% in 1 min and then kept constant for 4 min. The limit of quantification (LOQ) of the method was 0.01 g/kg.
For liquid chromatography mass spectrometry (LC/MS) analyses, the ESI interface was used in positive-ion mode, with the following settings: desolvation temperature of 600°C, capillary voltage at 0.80 kV, and sampling rate of 5 Hz. The mass spectrometer was operated in full-scan (600 to 800 m/z) and in single-ion recording (SIR) modes by monitoring the mass of beauvericin (784 m/z; elemental formula [MϩH] ϩ :C 45 H 57 N 3 O 9 ). MassLynx, version 4.1, mass spectrometry software was used for data acquisition and processing. The retention time for beauvericin was ϳ9.80 min. Beauvericin was quantified by measuring peak areas and comparing these values with a calibration curve obtained from standard solutions (48,67,68).
(iv) Determination and confirmation of fumonisin production. Ten grams of culture material was extracted with 15 ml of methanol-water (70:30, vol/vol) on an orbital shaker (150 rpm) for 60 min. Six milliliters of the extract was evaporated to dryness under a stream of nitrogen at 40°C. The residue was dissolved in 1.5 ml of acetonitrile-water (30:70, vol/vol), filtered with RC 0.2-m-pore-size filters (Phenomenex, Torrance, CA), derivatized, as described below, and quantified by HPLC and fluorescence detection (FLD). To derivatize a sample, 50 l of a sample extract was mixed with 50 l of o-phthaladehyde (OPA) by shaking for 50 s in the HPLC autosampler of an Agilent 1100 equipped with a binary pump and a column thermostat set at 30°C. The 100-l volume was injected by full-loop injection 3 min after addition of the OPA reagent for fumonisin analysis. The analytical column was a Symmetry Shield RP18 (4.6 by 150 mm, 5-m particle size; Waters) with a guard column inlet filter (0.5-m by 3-mm diameter; Postnova Analytics, Inc., Salt Lake City, UT). The mobile phase consisted of a binary gradient whose initial composition was 57% A (water-acetic acid, 99:1, vol/vol) and 43% B (acetonitrile-acetic acid, 99:1, vol/vol) and kept constant for 5 min. Solvent B was then linearly increased to 54% at 21 min, linearly increased again to 58% at 25 min, and finally kept constant for 5 min. The flow rate of the mobile phase was 0.80 ml/min. The fluorometric detector was set at an excitation wavelength of 335 nm and emission wavelength of 440 nm. Retention time for FB 1 was 17 min. The LOQ of the method was 0.01 g/kg.
Fumonisin B 1 was confirmed by UPLC with an Acquity QDa mass detector. The chromatographic separation was performed on an Acquity UPLC BEH C 18 column (2.1 by 100 mm; 1.7-m particle size) preceded by an Acquity UPLC in-line filter (0.2-m pore size). The temperature of the column was set at 50°C. The flow rate of the mobile phase was set at 0.4 ml/min. Solvent A was water, and solvent B was methanol, with both solvents containing 0.1% acetic acid. A gradient elution was used beginning with 90% A and 10% B. The gradient was changed from 10% to 50% solvent B in 10 min and kept constant for 4 min; it was linearly increased to 90% solvent B in 3 min, and then kept constant for 4 min. For column reequilibration, solvent B was decreased to 10% in 1 min and kept constant for 3 min.
For LC/MS analyses, the ESI interface was used in positive-ion mode, with the following settings: desolvation temperature of 600°C, capillary voltage of 0.80 kV, and sampling rate of 5 Hz. The mass spectrometer was operated in full-scan (100 to 800 m/z) and in single-ion recording (SIR) modes by monitoring the individual mass (FB 1 722.40 m/z). Retention time for FB 1 was 16 min. Empower 2 software (Waters) was used for data acquisition and processing. The LOQ was 0.01 g/ml for FB 1 (48,67,68).
Data availability. Genome sequences were deposited in GenBank under accession numbers JAAIFR000000000 for RC 298, JAAIFQ000000000 for RC 528, and JAAIFN000000000 for RC 2914.