Introduction

Fungi of the genus Aspergillus are greatly important in health, agriculture and food production. Common species such as Aspergillus flavus and A. fumigatus may be opportunistic animal and plant pathogens, particularly of immunocompromised humans and of crops such as corn, cotton and nuts (de Lucca 2007; Goldman and Osmani 2008; Amaike and Keller 2011; Mousavi et al. 2016). A. nidulans is an important model organism for eukaryotic genetics and cell biology (Nierman et al. 2005; Goldman and Osmani 2008). A. oryzae is used to ferment many popular Japanese fermented foods and beverages and in the production of heterologous proteins, while A. niger and A. terreus are used for industrial production of citric acid and lovastatin (Goldman and Osmani 2008; Machida et al. 2008; Bennett 2010). A. flavus is a cosmopolitan saprophyte and opportunistic pathogen that is a common producer of aflatoxins, potent hepatotoxins and carcinogens (Richard and Payne 2003). Due to food safety implications, there is significant interest in measuring the diversity of this species in different populations and identifying stable non-aflatoxigenic strains (Cotty 1990, 1994; Cotty and Mellon 2006; Moore et al. 2009), which have been reported to comprise 15.8–78.9% of natural fungal isolates (Hua et al. 2012; Jamali et al. 2012; Solorzano et al. 2014; Divakara et al. 2015) across the world. Most biocontrol A. flavus strains have aberrant aflatoxins biosynthetic gene clusters. For example, strain NRRL 21882 (“Afla-Guard”) does not produce aflatoxins, because of a large deletion of the aflatoxins biosynthesis gene cluster and part of the adjacent sugar utilization cluster (Chang et al. 2005). Strain NRRL 30797 (“K49”) has a premature TGA stop codon in the polyketide synthase gene aflC (Chang et al. 2012) and strain NRRL 18543 (“Af36”) has deletions of 17–61 bp in genes aflU, aflC, aflR and aflaV (Adhikari et al. 2016). The evolutionary processes responsible for the occurrence of non-aflatoxigenic A. flavus strains are still unclear. However, recombination events may have a major influence on genetic variability (Geiser et al. 1998; Olarte et al. 2012; Moore 2014). Moore et al. (2009) propose that lineage-specific recombination events among the late aflatoxins synthesis pathway gene (aflE to aflP, but especially aflW and aflX) led to gene losses and phylogenetic clades by which non-aflatoxigenic strains can be grouped; later generations can regain toxicity by recombination with chromosome III of an aflatoxigenic parent. Along the same lines, Chang et al. (2012) also claim that an ancestral recombination event between two morphotypes resulted in the progenitors of NRRL 18543 and NRRL 30797. The boom in genomic sequencing of filamentous fungi in the past decade not only permits comparisons of different species, it also allows us to delve into the genomic differences among strains within a species (Nierman et al. 2005). Thereby, we can link genotypes to desirable phenotypes, such as improved industrial metabolite production, lignocellulose degradation (Emtiazi et al. 2001; Raulo et al. 2016) and biological control of mycotoxigenic strains (Richard and Payne 2003). We aim to identify genetic variations in addition to aberrant aflatoxin biosynthesis clusters that may distinguish effective biocontrol A. flavus strains.

Genetic variability and diversity within a population positively correlate with population fitness and a better ability of the population to adapt to changes in the environment (Reed and Frankham 2003). Sources of such genotypic differences include mutations, polyploidy, homologous recombination, transposable elements (TEs) and individual migration in and out of populations. TEs are of special interest because of their contributions to gene regulation and genome structural variation (Argueso et al. 2008; Feschotte 2008; Bourque 2009; Klein and O’Neil 2018). Transposons are often over-represented in heterochromatic areas such as in and near the centromere (Pimpinelli et al. 1995; Round et al. 1997; Peterson-Burch et al. 2004; Tsukahara et al. 2012; Klein and O’Neil 2018). Their high copy numbers can cause them to comprise 0.02–29.8% of fungal genomes, 3–45% of metazoan genomes and up to 80% of plant genomes (Daboussi and Capy 2003; Bennetzen 2005; Hua-Van et al. 2005; Castanera et al. 2016).

There are two categories of TEs: class I retrotransposons that have a ‘copy and paste’ mechanism in which the TE is replicated and integrated in different locations of the genome via an RNA intermediate, and class II DNA transposons that ‘cut and paste’ themselves throughout a genome. Retrotransposons can be further classed as having long terminal repeats (LTRs), long interspersed nuclear elements (LINE) or short interspersed nuclear elements. Codon usage of retrotransposons in some fungal hosts is noted to be atypical of the organism, indicating horizontal gene transfer origins (Hansen et al. 1988; Oliver 1992). The Gypsy superfamily of LTR retrotransposons is commonly found in all eukaryotic phyla (Wicker et al. 2007), but the subfamily Tf1/sushi is only reported in fungi and a few vertebrates, suggesting limited horizontal transfer from fungi to vertebrates (Miller et al. 1999; Butler et al. 2001). In fungi, the effects of TEs on strain differentiation and speciation are not well-defined. However, they are suspected to have effects on the chromosomal rearrangements and duplications via transposon activity or recombination in Fusarium oxysporum (Hua-Van et al. 2000; Davière et al. 2001), Magnaporthe grisea (Thon et al. 2006), Phanerochaete chrysosporium (Larrondo et al. 2007) and Aspergillus spp. (Clutterbuck et al. 2008; Lind et al. 2017). TEs also can serve as a “fossil” record of genome differentiation. The strain-specific distributions of transposons Vader and ANiTa1 in industrial strains of A. niger indicate they were active during domestication, yielding different genome organizations (Braumann et al. 2007, 2008). However, TEs are not always a clear cause of genomic rearrangements (Davière et al. 2001).

Several TE families are detectable in Aspergillus, totaling 1.2–2.5% of the genomes of A. nidulans, A. fumigatus and A. oryzae (Clutterbuck et al. 2008). Three Aspergillus Tf1/sushi subfamily TEs are known: Afut in A. fumigatus (Neuveglise et al. 1996), AFLAV in non-aflatoxigenic A. flavus NRRL 6541 (Okubara et al. 2003; Hua et al. 2007) and AoLTR in A. oryzae RIB40 (Jin et al. 2014). All of these TEs are 6000–7799 bp sequences flanked by two 282–669 bp repeat sequences; have two open reading frames (ORF1 and ORF2) staggered by a − 1 frameshift; and encode a putative Gag-Pol polyprotein with domains for Gag capsid protein, aspartic proteinase, reverse transcriptase, RNase H, and integrase, which are universal structural features of active Gypsy retrotransposons (Wicker et al. 2007). A screen of over 50 A. flavus field isolates reveals that about half had incomplete gag and pol regions; none had a full AFLAV-like sequence (Hua et al. 2007). A. oryzae additionally has Crawler, a Mariner/Tc1-type transposon (Ogasawara et al. 2009). Uncharacterized degenerate retrotransposons dane1 and dane2 are also found in A. nidulans (NCBI GenBank accessions AF295689.1 and AF295688.1, respectively).

The newly reported non-aflatoxigenic strain WRRL 1519 is missing nearly 42 kb of the beginning of the 75 kb aflatoxins synthesis gene cluster and does not have a complete cyclopiazonic synthesis gene cluster (Yin et al. 2018). In this study, we continued bioinformatics analyses of WRRL 1519, comparing it to the aflatoxigenic strain NRRL 3357, and here we report that the genome organization of strain WRRL 1519 differed from than that observed in other inspected non-aflatoxigenic A. flavus strains. In an attempt to identify the cause of these differences, we further modeled the association of these differences with the increased number of repetitive elements in the WRRL 1519 strain.

Methods

Comparisons of predicted proteomes

Programs were used with default settings unless otherwise stated. Genomic sequences of the aflatoxigenic A. flavus strain NRRL 3357, and the non-aflatoxigenic A. flavus strains NRRL 21882, NRRL 30797, NRRL 18543 and WRRL 1519 were retrieved from the NCBI Genome database in January 2018 (Nierman et al. 2015; Weaver et al. 2017; Yin et al. 2018; Table 1). The annotated proteome of the aflatoxigenic A. flavus strain NRRL 3357 was available from the same database. We predicted the proteomes of the non-aflatoxigenic NRRL 21882, NRRL 30797, NRRL 18543 and WRRL 1519 using Augustus version 3.2.3 (Stanke et al. 2006) and GeneMark-ES (Suite version 4.33) trained on A. oryzae (Ter-Hovhannisyan et al. 2008). The resulting GTF files were combined into one annotation file using an in-house Python 2.7 script.

Table 1 General information about A. flavus strains and scaffold alignments to chromosomes of NRRL 3357
Full size table

Predicted protein sequences of A. flavus strains NRRL 3357, NRRL 21882, NRRL 30797, NRRL 18543 and WRRL 1519 were compared for homology (as indicated by high sequence similarity) by searching against the proteomes using HMMER 3.1b2 (Johnson et al. 2010) with an E value cutoff of 1−50. Using the annotated NRRL 3357 proteome, predicted protein sequences of the non-aflatoxigenic strains were assigned homology. Then, these assignments were used to map the relative locations and directions of the scaffolds along the eight putative chromosomes of NRRL 3357. Scaffolds with at least 20 homologous genes in agreement were visualized using Circos v 0.69 (Krzywinski et al. 2009). The method was validated by homology matching strain NRRL 3357 to itself (Supplementary Fig. 1A). MUMmer v 4.0.0 (Marçais et al. 2018) was additionally used to identify nucleotide macrosynteny between aflatoxigenic NRRL3357 and the other non-aflatoxigenic A. flavus strains.

Prediction of secondary metabolite gene clusters, repetitive sequences, candidate retrotransposons, regulatory sequences and secretory protein-coding genes

Since the aflatoxigenic cluster is partially missing in strain WRRL 1519, we searched for other differences in putative secondary metabolite gene clusters predicted by antiSMASH 3.0 (Weber et al. 2015) using Easyfig 2.2.2 (Sullivan et al. 2011). Clusters were compared between strains NRRL 3357 and WRRL 1519 to identify if other secondary metabolite clusters had large deletions. Regulatory sequences (upstream a maximum of 3 kbp and not overlapping an open reading frame) and secreted proteins were predicted using RSAT server (Nguyen et al. 2018) with the JASPAR 2018 core non-redundant transcription factor DNA-binding preferences matrix (Khan et al. 2018) and TargetP 1.1 with the non-plant option (Emanuelsson et al. 2007), respectively.

RepeatMasker version 4.0.7 (Smit et al. 2018) was used to identified all repetitive elements using the fungal-specific repetitive sequences database from RepBase volume 23, issue 5 (https://www.girinst.org/). Lower-scoring sequences in the same loci as another higher-scoring one were considered redundant and removed. We modeled random placement of all repetitive sequences in the WRRL 1519 genome using a custom Python script: weighted by scaffold size, the script randomly assigned the 883 repetitive sequences to genomic locations on a scaffold at least one nucleotide larger in size. We then counted the number of randomized loci within a specified range of nucleotides bordering predicted protein-coding genes, regulatory sequences, other repetitive sequences and secretory protein-coding genes. Ten thousand permutations were performed for each instance. The randomized model was compared to the actual counts of repetitive sequences within the borders of the sequences of interest. Genomic densities and locations of sequences of interest were visualized using Circos.

Reverse transcriptases and retrotransposons identified from A. flavus sequences and deposited in NCBI GenBank were queried against the genomic sequences of inspected strains using HMMER with an E value cutoff of 1−200. Additional LTR sequences were search with a 1−50 E value threshold. Sequence similarities of retrieved homologous sequences were visualized using MUSCLE 3.8.31 (Edgar 2004) and BOXSHADE 3.21 (https://embnet.vital-it.ch/software/BOX_form.html). Candidates were queried against the NCBI non-redundant protein database (O’Leary et al. 2016; accessed March 2018) using NCBI BLAST (Altschul et al. 1990; https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE=Nucleotides&PROGRAM=blastn&PAGE_TYPE=BlastSearch&BLAST_SPEC=). The gene environment of the eleven retrotransposon-like sequences in WRRL 1519 was characterized based on alignment to annotated NRRL 3357 genes.

Results

Strain WRRL 1519 was different in genomic organization from other A. flavus strains

Most of the predicted protein in all non-aflatoxigenic strains were assigned homology to at least one NRRL 3357 protein (Table 1). However, 512 NRRL 3357 genes were commonly not assigned homology to any of the non-aflatoxigenic strains (Fig. 1). The latter all corresponded to short sequences, 50–91 amino acids in length, that were mostly annotated as “conserved” and/or “hypothetical”. All could be identified as homologous to something if the E value threshold was lowered to 1−10. There were 1, 902 NRRL 3357 proteins that did not match to themselves when queried against the NRRL 3357 proteome or to anything in the predicted proteomes of the biocontrol strains, mostly consisting of conserved and hypothetical proteins, as well as some transcription factors, transmembrane transporters, kinases, enzymes and domain consensus motifs.

Fig. 1
figure 1

Venn diagram of NRRL 3357 proteins not found to have homology to predicted proteins of the non-aflatoxigenic strains. The image was drafted with the venn R package version 1.6 (Dusa 2016; R Core Team 2015) and edited with Inkscape version 0.91 (The Inkscape Team 2015)

Full size image

The similarities of predicted protein-coding genes between NRRL 3357 and non-aflatoxigenic strains were examined to map the respective scaffolds to the chromosomal arms of NRRL 3357. The 16 largest scaffolds of the A. flavus NRRL 3357 genome assembly comprised the 16 arms of the eight chromosomes of A. flavus. The largest 16 or 17 scaffolds of strains NRRL 21882, NRRL 30797and NRRL 18543 were homologous to these chromosomal arms (Supplementary Fig. 1B–D). Only strain NRRL 21882 did not have a clear match to the p arm of chromosome VII (NRRL 3357 chromosome EQ963487.1). The genome sequence of WRRL 1519 had 127 scaffolds of which we were able to align 84 to the nuclear genome of NRRL 3357 (Fig. 2a). Scaffold 98 was the mitochondrial genome. Scaffolds 59, 72, 73, 82, 84, 88, 90–92 and 94–127 could not be placed confidently as they had too few homologous genes or unclear alignments. All non-aflatoxigenic strains had a few genes that matched to those the 17th largest scaffold of NRRL 3357 EQ963488.1. Predicted protein-coding genes from nucleotide 37,001 to 51,632 of scaffold 65 from WRRL 1519 mapped to EQ963488.1. However, the gene sequences from 57,085 to 134,489 mapped to EQ963475.1, so this scaffold was assigned to chromosome III. Similarly, one 2636 nucleotide gene in scaffold 0 was homologous to a gene in extrachromosomal scaffold EQ965297.1, but most of the WRRL 1519 sequences aligned with NRRL chromosomal scaffolds EQ963473.1 and EQ963481.1. For the other non-aflatoxigenic strains, the 17th or 18th largest scaffold of the genomes matched to EQ963488.1. The orders and orientations of the placed scaffolds relative to the NRRL 3357 chromosomal scaffolds strain are shown in Fig. 2, Supplementary Table 1 and Supplementary Fig. 2. Results from the protein similarity searches between NRRL 3357 and WRRL 1519 are shown in Supplementary Table 2. MUMmer results with nucleotide-based similarity searches were in accordance in that the strains NRRL 21882, NRRL 30797 and NRRL 18543 were noticeably more similar in genome organization to NRRL 3357 than WRRL 1519 (Supplementary Fig. 3).

Fig. 2
figure 2

Scaffold ordering of strain WRRL 1519 to the chromosomes of NRRL 3357. Scaffolds comprising the chromosomal arms of NRRL 3357 are drawn in 16 different colors on the right and are labeled as in the genome assembly (see Supplementary Table 1). The outer track enumerates chromosome nucleotide length in kilobase pairs, and protein-coding gene density is shown by a gray histogram on the inner track. On the left, scaffolds of WRRL 1519 are connected to the chromosomes by lines representing high sequence similarity between two predicted protein-coding genes. Used scaffolds have at least 20 genes homologous to one NRRL 3357 scaffold. Beige scaffolds have deposited nucleotide sequences that run in the same direction as the deposited sequence of the aligned NRRL 3357 chromosomal arm; black ones run in the opposite direction

Full size image

Most of the scaffolds of the non-aflatoxigenic strains aligned well with the putative chromosomal arms of aflatoxigenic NRRL 3357. Extrachromosomal differences were defined as a gene in a particular non-aflatoxigenic scaffold having best homology to a gene in a NRRL 3357 chromosomal arm to which the scaffold is not aligned (Table 1). For example, scaffold 30 of strain WRRL 1519 is homologous to a large region of chromosomal arm EQ963477.1. However, the best overall match for this scaffold was to EQ963472.1. These differences could be the results of lower E value assignment to non-homologous genes by coincidence, to true extrachromosomal paralogs or to true homologs with different relative chromosomal placements. For most of these extrachromosomal differences, the assigned NRRL 3357 homolog was the only hit retrieved from the PHMMER searches.

Strains NRRL 21882, NRRL 30797 and NRRL 18543 had 134 extrachromosomal mismatches common among them. Strain WRRL 1519 shared 122 of these mismatches. The mismatched genes were most frequently annotated as transcription factors, transporters/permeases, non-ribosomal peptide synthetases and polyketide synthases. There were five large sections of extrachromosomal differences between WRRL 1519 and NRRL 3357 clearly visible in both Fig. 2 and Supplementary Fig. 3: between WRRL 1519 scaffold 3 and NRRL 3357 scaffold EQ963484.1 (chromosome VI, p arm), scaffold 30 and scaffold EQ963477.1 (chromosome V, q arm), scaffold 15 and scaffold EQ963482.1 (chromosome IV, p arm), scaffold 40 and scaffold EQ963479.1 (chromosome VI, q arm), and scaffold 13 and scaffold EQ963477.1 (chromosome V, q arm). The mismatches were not associated with significant differences in the structures of predicted secondary metabolite gene clusters (Supplementary Table 3). However, two WRRL 1519 gene clusters aligned to NRRL 3357 chromosomes EQ963476.1 (chromosome VII, q arm) and EQ963486.1 (chromosome VIII, p arm) had large deletions (Fig. 3). The former is predicted to yield a terpenoid non-ribosomal peptide product, the latter a non-ribosomal peptide product.

Fig. 3
figure 3

Alignments of predicted secondary metabolite gene clusters of strains NRRL 3357 (top) and WRRL 1519 (bottom) between a chromosome EQ963476.1 and scaffold 11 and b chromosome EQ963486.1 and scaffold 7. Percent identity by tblastx between subsequences is shown as indicated by the key. Genes with predicted functions in the NRRL 3357 proteome are green; others are purple. Automatic gene annotation of WRRL 1519 may have resulted in incorrect start and stop positions

Full size image

Strain WRRL 1519 had an increased number of repetitive elements

Repetitive elements may affect genome structure, so we searched for such sequences using RepeatMasker and HMMER. RepeatMasker revealed that strain WRRL 1519 (886 elements comprising 1.54% of the genome) had an increased number of fungal-specific repetitive elements relative to the other strains (NRRL 3357—603 elements, 0.43% of the genome; NRRL 21882—477 elements, 0.28% of the genome; NRRL 30797—533 elements, 0.36% of the genome; NRRL 18543—533 elements, 0.34% of the genome) (Fig. 4). Gypsy and Mariner/Tc1-like sequences were the most populous in all strains. Among the repetitive sequences in the Gypsy superfamily, AFLAV-like repeats were masked nearly 6 times more frequently in the WRRL 1519 (59 occurrences as opposed to 6–10 in other A. flavus strains).

Fig. 4
figure 4

Frequency of repetitive sequence elements in the A. flavus genomes. A total of 603, 477, 533, 533 and 886 non-redundant repetitive sequences were detected in strains NRRL 3357, NRRL 21882, NRRL 30797, NRRL 18543 and WRRL 1519, respectively

Full size image

We attempted to identify complete putative reverse transcriptase and retrotransposon nucleotide sequences by querying specific sequences against the genomes of the inspected strains. Eleven candidate retrotransposons and six reverse transcriptases were found in WRRL 1519. In contrast, strains NRRL 3357, NRRL 21882, NRRL 30797 and NRRL 18543 had a maximum of one detected retrotransposon (NCBI GenBank accession AFLA_063980) and/or five reverse transcriptases (accessions AFLA_001110, AFLA_018200, AFLA_053840 and AFLA_114250). Genomic sequences from non-aflatoxigenic strains NRRL 21882, NRRL 30797, NRRL 18543 and WRRL 1519 did not share homology with the one queried LINE-1 retrotransposon (accession AFLA_063980 found in NRRL 3357). Only WRRL 1519 had sequences homologous to AFLAV (accession AY485786.2). The length of this query sequence was 7779 bp; most of the candidate retrotransposons were at least 6000 bp (Supplementary Table 4). The candidate retrotransposons in strain WRRL 1519 most similar to AFLAV were scaffold_113, scaffold_8b and scaffold_18a (Supplementary Fig. 4), suggesting that most of the candidates were derived from other non-AFLAV-like TEs, or the insertions were considerably older and have accumulated more mutations. Additionally, five of the eleven candidates were located in putatively pericentromeric scaffolds; most of the AFLAV sequences detected by RepeatMasker or HMMER were located on scaffolds predicted to be near the centromeres or the telomeres (Supplementary Fig. 2). Three of the candidates had canonical open reading frames at least 300 bp long: scaffold_8b (303 bp), scaffold_1 (390 bp) and scaffold_113 (6000 bp). According to NCBI BLAST results, the latter two open reading frames contained sequences likely to be related to Gag-Pol polyproteins. Moreover, BLAST results against the NCBI database indicated that most of the candidates were highly similar to subsequences on scaffolds SC005 and SC009 of A. oryzae RIB40 (E value < 1−247), corresponding to the 6000 bp retrotransposon AoLTR. Separate from the LTRs reported in Table 5, several sequences similar to both the 5′ and 3′ LTRs of AFLAV were identified in WRRL 1519 scaffolds 1 (two instances), 8 (four instances), 18, and 38. None were found in the other A. flavus strains.

Repetitive sequences were associated each other, but not with the differences in genome organization

The genetic neighborhood of the candidate retrotransposons included genes that were largely unannotated, except for candidate scaffold_22 which was surrounded by genes annotated as citrate synthase, transcription factor, sugar transporter and ABC multidrug transporter. Candidate retrotransposons on scaffolds 0, 1, 8 and 89 were in areas with chromosomal mismatches and genes that did not have homology to NRRL 3357 genes. Candidate scaffold_0 was located at the interface of where scaffold 0 stops matching to mostly proteins in chromosome EQ963481.1 (chromosome II, p arm) and begins matching those in EQ963473.1 (chromosome II, q arm). Candidate scaffold_1 neighbors a single extrachromosomal mismatch to EQ963480.1 (chromosome V, p arm), by a scaffold that mostly matches EQ963478.1 (chromosome II, q arm). This mismatch adds a copy of a kinesin family protein-like coding gene. AFLAV-like sequences on scaffold 8 were located in a region (nucleotides 155,663 to 341,186) where only 10 out of 44 predicted genes had homology to those in NRRL 3357. The AFLAV candidate on scaffold 89 was located at the end of a mismatch to a region of EQ963481.1 (chromosome II, p arm) NRRL 3357 that is not represented in other WRRL 1519 scaffolds.

Noting that five of the AFLAV candidates were within 100,000 bp of apparent extrachromosomal mismatches, we suspected that TE activity may be responsible for chromosomal shifts. We inspected the genomic loci of all extrachromosomal predicted protein-coding genes, regulatory sequences, repetitive elements and secretory genes and performed permutation tests to see if the repetitive elements were generally associated with any of these sequences of interest. Over half of the WRRL 1519 repeat sequences were clearly clustered with one another within 10,000 bp (Figs. 5, 6a). However, permutations in which the loci of repetitive sequences were randomized did not reveal significant associations with extrachromosomal mismatches for any of the strains (Fig. 6b).

Fig. 5
figure 5figure 5

Genomic locations of genes, regulatory and repetitive sequences and secondary metabolite gene clusters on NRRL 3357 (left) and WRRL 1519 (right) scaffolds constituting chromosomes a I–III and b IV–VIII. Densities of predicted protein-coding genes, regulatory sequences, repetitive sequences and secretory protein-coding genes are represented by the gray (max height = 19), green (max height = 231), red (max height = 14) and purple (max height = 9) histograms, respectively. Bin size (25,000 bp) is the same for all histograms. The innermost layer depicts genomic locations of secondary metabolite gene cluster colored by product type (pink, polyketide /non-ribosomal peptide; orange, siderophore; yellow, terpene; blue, indole; black, other). Scaffold length is tracked in kilobase pairs. (Color figure online)

Full size image
Fig. 6
figure 6

Association of repetitive sequences with a other repetitive sequences and b extrachromosomal mismatches. The percentage of total repeats for a strain within a specified range of the borders of any repetitive sequences or extrachromosomal mismatch in the strain’s genome detected from the proteomic alignments is shown. Repeats from the strains and the randomized model are indicated by colored shapes. a The noticeable deviation of the A. flavus strains from the randomized pattern indicates that the repetitive elements were non-randomly associated with one another. b Repeat sequences were not significantly associated with extrachromosomal mismatches. Repeats similarly appeared randomly associated with protein-coding genes, regulatory sequences and secondary metabolite gene clusters (data not shown)

Full size image

Discussion

The first published genome of A. oryzae was sequenced to 9X coverage depth, initially yielding six scaffolds and ten contigs (24 contigs total) organized into eight chromosomes by Southern hybridization to chromosomal probes and fingerprinting; the assembly was validated by cloning and optical mapping (Machida et al. 2005). A. oryzae and A. flavus are closely related to A. flavus. Therefore, when aflatoxigenic A. flavus NRRL 3357 was sequenced, the 16 largest scaffolds (out of 79 scaffolds) were organized into eight chromosomes by sequence comparison with the A. oryzae chromosome map. The overall genomic organizations between the species were similar except for a large translocation between chromosomes II and VI (Payne et al. 2006, 2008). In turn, the genome of Aspergillus parasiticus SU-1 was also mapped to the 16 scaffolds of A. flavus by sequence similarity (Linz et al. 2014).

Thus, using the alignment of aflatoxigenic A. flavus NRRL 3357, we were able to organize the genomic scaffolds of the four non-aflatoxigenic strains into hypothetical chromosomes. Most of the genome assemblies of the non-aflatoxigenic strains had at least one scaffold matched to a putative NRRL 3357 chromosomal arm. However, the p arm of the chromosome VII was not found in the genome assembly of NRRL 21882. None of the scaffolds had a similar size to the arm (~ 350 kbp) and the only four matched protein-coding genes were on four different scaffolds. However, this chromosomal arm was found be represented by scaffold NKQQ01000002.1 in an older genome assembly of NRRL 21882 (NCBI genome assembly ID GCA_002217635.1), so the newer assembly simply may be incomplete. We were able to align most of the WRRL 1519 genome assembly to that of NRRL 3357, providing a guide to future improvement of the current genome assembly. Our current results are purely computational and the scaffold ordering should be verified by deep sequencing.

In addition to building a better hypothetical genome assembly, we were able to compare the organization of protein-coding gene loci in the genomes of NRRL 3357 and WRRL 1519. There was a notably higher number of extrachromosomal mismatches between NRRL 3357 and WRRL 1519 than between NRRL 3357 and any other tested non-aflatoxigenic strains. The biological significance, if any, of these apparent chromosomal rearrangements is not clear. None of the secondary metabolite biosynthesis gene clusters at those locations were noticeably different between strains. However, single genes or regulatory sequences may be missing or disrupted due to the difference in genome organization that we have not detected. The apparent deletions in two predicted biosynthetic gene clusters of WRRL 1519 were not associated with extrachromosomal mismatches, and likely indicate a difference in the secondary metabolite repertoires of NRRL 3357 and WRRL 1519.

The cause of the extrachromosomal mismatches is also not clear. There was a marked increase in transposon-like sequences in WRRL 1519. TEs can cause insertional mutations in fungi and induce differences in genome structure among strains, driving evolution (Nishimura et al. 2000; Braumann et al. 2008; Lind et al. 2017). TEs may land within an open reading frame or regulatory sequence. TE activity may be actively repressed by host-directed repeat-induced point mutations and such mutations may “leak” over to neighboring genes (Irelan et al. 1994; Fudal et al. 2009). Generally, TEs may cluster preferentially near genes encoding effectors (secreted cysteine-rich proteins that modulate host-pathogen interactions), increasing the evolutionary speed of genes involved in pathogenicity; likewise TEs cluster near the immune genes of plants (Raffaele and Kamoun 2012; Dong et al. 2015; Seidl and Thomma 2017). In Aspergillus, some TEs are associated with gene clusters (Lind et al. 2017).

Transposon insertion and activity can potentially affect virulence of phytopathogenic fungi. The location of a class I Mariner/Tc1 transposon in the promotor or coding sequence of an avirulence gene allows M. oryzae to overcome gene-for-gene resistance (Kang et al. 2001; Zhou et al. 2007). Xu et al. propose that frequent transposon activity is responsible for the evolution of a pathogenic M. oryzae ancestor to the endophytic Harpophora oryzae (Xu et al. 2014). Active and inactive transposons from several plant pathogenic fungi including F. oxysporum (Anaya and Roncero 1996; Daboussi 1997; Mes et al. 2000; Rep et al. 2005), Verticillium dahlia (Amyotte et al. 2012), M. oryzae (Ikeda et al. 2001; Chadha and Sharma 2014) and Pseudocercospora fijiensis (Chang et al. 2016) are associated with increased genome size, genome instability and disease aggressiveness. Still active transposons can often be stimulated by stress conditions (Anaya and Roncero 1996; Ikeda et al. 2001; Ogasawara et al. 2009; Amyotte et al. 2012; Chadha and Sharma 2014).

Scientific publications make special note of their appearance often found near effectors, virulence genes, and gene clusters (Nishimura et al. 2000; Braumann et al. 2008; Raffaele and Kamoun 2012; Dong et al. 2015; Lind et al. 2017; Seidl and Thomma 2017). We have shown that repetitive sequences are significantly likely to be within 10,000 bp of one another, may be associated with chromosomal differences and that genome-wide associations with extrachromosomal mismatches and other sequences of interest are not obviously different from what could be randomly expected, even in biocontrol strain WRRL 1519 which exhibited a noticeably different genome organization. Payne et al. (2006, 2008) note that the frequencies of three types of TEs are consistently greater in the atoxigenic A. oryzae than in A. flavus, implying that TEs may present a method by which some non-aflatoxigenic strains may arise and differentiate from their aflatoxigenic ancestors. Chang and Ehrlich (2010), reporting that NRRL 3357 has low copy numbers of elements Tao1, Crawler and AFLAV, further state that mobile elements may contribute to the differentiation of A. flavus and A. oryzae. In light of our results, we also hypothesize that transposons may play some role in shaping the genomes of some non-aflatoxigenic A. flavus strains that can be used for biocontrol of aflatoxins. Whether transposons may lead to the nonrandom natural evolution of fungal strains in nature is an intriguing question and is one we are pursuing in future work.