Novel Mitoviruses and a Unique Tymo-Like Virus in Hypovirulent and Virulent Strains of the Fusarium Head Blight Fungus, Fusarium boothii

Hypovirulence of phytopathogenic fungi are often conferred by mycovirus(es) infections and for this reason many mycoviruses have been characterized, contributing to a better understanding of virus diversity. In this study, three strains of Fusarium head blight fungus (Fusarium boothii) were isolated from Ethiopian wheats as dsRNA-carrying strains: hypovirulent Ep-BL13 (>10, 3 and 2.5 kbp dsRNAs), and virulent Ep-BL14 and Ep-N28 (3 kbp dsRNA each) strains. The 3 kbp-dsRNAs shared 98% nucleotide identity and have single ORFs encoding a replicase when applied to mitochondrial codon usage. Phylogenetic analysis revealed these were strains of a new species termed Fusarium boothii mitovirus 1 in the genus Mitovirus. The largest and smallest dsRNAs in Ep-BL13 appeared to possess single ORFs and the smaller was originated from the larger by removal of its most middle part. The large dsRNA encoded a replicase sharing the highest amino acid identity (35%) with that of Botrytis virus F, the sole member of the family Gammaflexiviridae. Given that the phylogenetic placement, large genome size, simple genomic and unusual 3′-terminal RNA structures were far different from members in the order Tymovirales, the virus termed Fusarium boothii large flexivirus 1 may form a novel genus and family under the order.


Introduction
The number of extant fungal species is currently estimated at up to 1.5 million. Viruses that infect fungi are known as mycoviruses or fungal viruses and are ubiquitous across the kingdom Fungi [1]. Partitiviruses, and mitoviruses in particular, are omnipresent. Interestingly, mycoviruses were often found to have unique features such as genomic structures, virion structures, anti-viral defence strategies, etc. that differ from major plant and animal viruses [2][3][4]. Most mycoviruses infect fungal hosts asymptomatically and the remainder rarely exert a positive or negative impact on fungal host traits such as virulence, asexual/sexual spore production, pigmentation and growth [5][6][7]. Virocontrol, a method of biological control using mycoviruses that attenuate the virulence of phytopathogenic fungi (conferring hypovirulence), has attracted the attention of researchers for the last few decades [8][9][10][11], Tymovirales that has no capsid protein (CP) [41,42]. Recently, Rhizoctonia solani flexivirus 1 (RsFV-1, unclassified deltaflexivirus-like) was reported to have the largest genome in the order and its unique genomic organisation broadened our perception of genome plasticity within the order [38]. In addition, two mycoviruses, Fusarium graminearum mycotymovirus 1 (FgMTV1, unclassified tymovirus-like) and Fusarium graminearum deltaflexivirus 1 (FgDFV1, Deltaflexiviridae), have been isolated from the FHB pathogen, F. graminearum [20,27]. Biological assays demonstrated that FgMTV1 infection does not alter the host virulence, but decreases the growth rate and DON production. Finally, although oyster mushroom spherical virus (OMSV) was earlier reported to be a tymo-like mycovirus [44], its taxonomic placement is inconclusive till date.
This study characterised a novel member of Tymovirales, potentially forming a new genus or family, and members of a new species in the genus Mitovirus in the family Narnaviridae. These virus classes were identified for the first time in Fusarium boothii strains from the African continent. It is noteworthy that the tymo-like virus possesses the largest genome size (over 12.5 kb) among the currently reported Tymovirales members. This study contributes to our understanding of Tymovirales virus and mitovirus diversity and evolution.

Fungal Strains and Culture
F. boothii strains (Ep-BL13, Ep-BL14 and Ep-N28) were isolated from wheats with the FHB symptoms in Hadiya region of southern Ethiopia. A Japanese isolate of F. graminearum sensu stricto (s.str.) strain was also used as a reference. All strains were maintained at 28 • C on potato dextrose agar (PDA, BD Difco Laboratories, Detroit, MI, USA) and synthetic low nutrient agar (SNA) and stored at 4 • C in the dark. Mycelia were cultured in potato dextrose broth (PDB, BD Difco Laboratories, Detroit, MI, USA) liquid medium at room temperature for 4-7 days for nucleic acid extractions.

Biological Assessment of FGSC Strains
Growth rate and colony morphology. Mycelial plugs were transplanted on PDA and SNA plates (three replicates) and incubated at 20 • C for 1 week. The longest and shortest diameters of each colony were measured 3-7 days post-transplanting and the mean diameter was used for the calculation of colony area.
Pathogenicity test. Conidial spores (1 × 10 5 spores/mL) suspended in a 1% (v/v) Triton solution were used for inoculation, and the FHB-sensitive wheat cultivar Apogee was grown at 27 • C in a chamber (KG-50HLA; Koito Electric Industries, Yokohama, Japan) to avoid possible release of inocula into the environment [46,47]. Briefly, 10 µL of spore suspension was injected into wheat spikelets, and the number of diseased spikelets was counted at 15 days post-inoculation. Alternatively, the spore suspension was sprayed three times (500 µL/push) to a wheat head, and disease severity was evaluated according to Ban and Suenaga's criteria (index, mild to severe; 0, 5,10,20,30,50,60,80, 100) at 15 days post-inoculation [48]. Inoculated plants were placed in a transparent humid box for one day (injection method) or two days (spray method) to promote fungal germination, prior to the maintenance in the chamber.

dsRNA Extraction and cDNA Library Construction
dsRNA fraction was extracted from fungal mycelia by conventional cellulose column chromatography. Filtrated mycelia were homogenised in the presence of liquid nitrogen, and total RNA fractions were obtained by treatment with one round each of phenol, phenol-chloroform-isoamylalcohol (25:24:1), and chloroform-isoamylalcohol (24:1) extraction. dsRNAs were further isolated from the total RNA fractions by using Cellulose Powder C (Advantec, Tokyo, Japan). To eliminate fungal chromosomal DNA and ssRNA species, these fractions were treated with RNase-free RQ1 DNase I (Promega, Madison, WI, USA) and S1 Nuclease (Promega). Conventional cDNA library construction was performed for viral sequence determination. The dsRNAs extracted from F. boothii Ep-BL13, Ep-BL14 and EP-N28 and the 3 kbp-and 2.5-kbp dsRNA bands were gel-purified using a Zymoclean™ Gel RNA Recovery kit (Zymo Research, Irvine, CA, USA). cDNA libraries of each dsRNA fragment were constructed using a non-PCR and a PCR-based method. After denaturation of dsRNA templates at 99 • C for 5 min, cDNA was synthesised using ReverTra Ace-α-(Toyobo, Osaka, Japan) with an adapter-tagged primer (5 -CCTGAATTCGGATCCTCCNNNNNN-3 ). The resulting cDNA fragments were purified from an agarose gel with a size exclusion (1.0-2.0 kbp) and amplified by PCR with an adapter primer (5 -CCTGAATTCGGATCCTCC-3 ). The PCR conditions are as follows: 94 • C (2 min), then 1 cycle at 94 • C (2 min)/65 • C (1 min)/72 • C (1 min), followed by 35 cycles at 94 • C (40 s)/52 • C (30 s)/72 • C (3 min), and a final extension at 72 • C for 7 min. DNA fragments were cloned into the pGEM-Teasy (Promega) or pCR-Blunt (Thermo Fisher, Waltham, MA, USA) cloning vectors and used for the transformation of Escherichia coli strain DH5α for Sanger sequencing analyses.
2.5. RNA Seq, Reverse Transcription PCR and Rapid Amplification of cDNA Ends RNA seq. The F. boothii Ep-BL13 virome was established on an Illumina-Hiseq 2500 system (Illumina, San Diego, CA, USA) by a pair-end sequencing run (2 × 100 bp) using 10 ng of dsRNA as a template. cDNA and library construction were outsourced to BGI Japan. Fastqc version 0.11.5 software (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) was used to check the quality of the read data [51]. Reads were filtered using Sickle version 1.33 software (https://github.com/najoshi/ sickle) by requiring an average Phred quality (Q score) of at least 30 and to only retain reads of 20 nucleotides (nt) or greater in length [52]. De novo assembly was carried out using SPAdes version 3.11.1 software (http://cab.spbu.ru/software/spades/) at the default settings [53].
Gap-filling RT-PCR. Target cDNAs were produced using ReverTra Ace and a gene-specific primer set from given dsRNA fragments and amplified with DNA polymerases such as PrimeSTAR (Takara, Shiga, Japan), KOD FX Neo (Toyobo) or GO-Taq (Promega).

Phylogenetic Analysis
Multiple amino acid or nucleotide sequence alignments were constructed using MUSCLE [56] and with the MEGA X [57]. Phylogenetic analysis was conducted by using PhyML ver. 3.0 (http: //www.atgc-ontpellier.fr/phyml/) with the best-fit models suggested by Smart Model Selection in PhyML (http://www.atgc-montpellier.fr/sms/) [58], and the results were visualised with FigTree and further enhanced in Adobe Illustrator.

Detection of dsRNAs in F. boothii Strains
The fungal strains Ep-BL13, Ep-BL14 and Ep-28 were isolated from wheats exhibiting FHB in Ethiopia by using Fusarium-selecting medium, which were found to be dsRNA-positive. Agarose gel electrophoresis of dsRNA fractions extracted from these strains showed a 3 kbp common to all three strains and additional >10 and 2.5 kbp bands only in the Ep-BL13 strain ( Figure 1A), thereby suggesting infections by unknown RNA mycoviruses. These fungal strains were identified as F. boothii based on comparison of partial TEF1-α and ITS sequences (data not shown) and by phylogenetic analysis of partial TEF1-α sequences ( Figure S1).

Biological Characteristics of F. boothii Strains
Colony morphology of the F. boothii strains on nutrient-rich (PDA) and -poor (SNA) solid media varied as shown in Figure 1B,C. The Ep-BL13 strain grew most slowly on PDA, followed by Ep-BL14 and then Ep-28 ( Figure 1B,D). The strains also exhibited different phenotypes on SNA, with Ep-BL13 and Ep-N28 growing faster and slower than those on PDA, respectively ( Figure 1C). The pathogenicity of the three F. boothii isolates and of a reference F. graminearum s.str. strain on wheat head was investigated with injection and spray methods. Plants inoculated with F. boothii Ep-BL14 and Ep-28 exhibited typical FHB symptoms similar to that of the plant infected by F. graminearum s.str., whereas the Ep-BL13 showed apparently weak virulence ( Figure 1E,F). The effects of the dsRNA elements on host fungal growth and pathogenicity to wheats are yet to be determined.

Analysis of Viral Sequences: A Tymo-Like Virus
The genome organisation of the coding strand of the largest dsRNA segment harboured in Ep-BL13 is shown in Figure 2A. The complete genomic sequence of the virus was 12,579 nt in length and the G + C content of the dsRNA is 64.93% (accession No. LC425115). A BLASTX search of the whole nucleotide sequence showed that the highest amino acid (aa) identity (35%) was to the replicase of BVF, which is the sole member of the family Gammaflexiviridae (Table 1). Based on this result, the virus was tentatively named Fusarium boothii large flexivirus 1 (FbLFV1). The whole genome of FbLFV1 encompasses a single large ORF putatively encoding a polypeptide of 4070 aa (162-12,374 nt) with a calculated molecular mass of 443.6 kDa. However, BVF, the closest virus to FbLFV1 in a BLAST search, possesses two ORFs encoding a putative replicase (a potential readthrough product) and a putative CP. Methyltransferase (Met), helicase (Hel) and RNA-dependent RNA polymerase (RdRp) domains of the FbLFV1 polypeptide showed low aa identities (22-36%, 20-38% and 25-37%, respectively) to those of representative viruses in the order Tymovirales and related mycoviruses, as shown in Table 1. Three replicase domains, Met, Hel and RdRp, are located in this order, from N to C terminus, and this is typical of the "alphavirus-like" superfamily that includes many other plant-infecting viruses ( Figure 2A) [59]. There were two regions in the polypeptide with similarity to a PHA03247 domain (N-terminal part, E-value: 3.94 × e −6 ; C-terminal part, E-value: 5.01 × e −3 ), which was found in the large tegument protein UL36 of herpesviruses (a large dsDNA virus), although the function of PHA03247-like domains in FbLFV1 is unknown.

Analysis of Viral Sequences: A Tymo-Like Virus
The genome organisation of the coding strand of the largest dsRNA segment harboured in Ep-BL13 is shown in Figure 2A. The complete genomic sequence of the virus was 12,579 nt in length and the G + C content of the dsRNA is 64.93% (accession No. LC425115). A BLASTX search of the whole nucleotide sequence showed that the highest amino acid (aa) identity (35%) was to the replicase of

Molecular Phylogenetic Analysis of FbLFV1
Amino acid sequences of whole FbLFV1 replicase and representative mycoviruses in the order Tymovirales, including taxonomically unassigned members, were aligned and the Met, Hel and RdRp domains were extracted (Table 1 and Figure 3). These three domains were found to be well conserved in the deduced protein sequence of FbLFV1 replicase. The evolutionarily conserved relationship of FbLFV1 to the Tymovirales members was characterised by maximum likelihood (ML) phylogenetic analyses based on the alignments of each of the three domains ( Figure 4). As expected from BLAST analyses, the FbLFV1 Met Hel and RdRp domains were distantly associated with those of viruses in the order Tymovirales, with relatively close relationship to a gammaflexivirus ( Figure 4A-C). The concatenate sequence of these domains finally revealed that FbLFV1 was placed independently from the clades of established families in the order Tymovirales, although the branching support value was not sufficiently high in the ML tree ( Figure 4D). Based on these phylogenetic analyses, FbLFV1 is expected to belong to a novel virus species "Fusarium boothii large flexivirus 1" that may form an independent genus, or even a new virus family, in the order Tymovirales. The initiation codon at 162-164 nt of the large ORF was in a favourable context, CACCAUGGa, for filamentous fungi, although another potential initiation codon at 27-29 nt appeared in frame to the ORF with a weak context uugCAUGcG (capital and small letters represent good and poor residues in the context, respectively; underlined letters are the putative initiation codon; Figure 2B) [60][61][62].

Presence of a Defective Segment of FbFV1 in the Ep-BL13
The cDNA-library-based sequencing of the approximately 2.5 kbp dsRNA revealed that the segment was of a defective RNA (D-RNA) of FbLFV1. The total length of the dsRNA was 2408 bp, and the positive-sense RNA retained an ORF encoding polypeptide with a potential molecular mass of 76.8 kDa (716 aa; Figure 2A) (accession No. LC425116). This occurred because of the loss of aa 303-3693 of the FbLFV1 large polypeptide (removal of FbLFV1 genome positions 1066-11,235 in-frame) and the addition of 36 aa because of a frame-shift from a single nucleotide deletion at −5 nt position relative to the original termination codon, leading to a new downstream stop codon. Since the RNA sequence was almost identical to the corresponding region of FbLFV1 and the junction of the upstream/downstream parts was uniform, this D-RNA of FbLFV1 might have arisen very recently, but at least before the isolation of Ep-BL13. The segment has the potential to facilitate viral replication because the coded protein retained partial RdRp motifs (122 aa residues of C-terminal end; Figure 2A); however, it is also possible that it may interfere with replication. As BVF is reported to carry similar D-RNA species with unknown function [39], it is interesting to investigate the significance of D-RNAs in life cycles of mycoviruses in the order Tymovirales.

Predicted RNA Structure of the Terminal Regions of FbLFV1 Genome
The UTRs of RNA viruses generally carry important cis-elements that form secondary/tertiary structures and/or distant base-pairing. In many cases, these are essential for encapsidation, transcription, translation and replication. To investigate whether FbLFV1 contains such structures, UTR sequences were subjected to secondary structure prediction using mfold software. The 5 UTR of FbLFV1 did not consistently exhibit stem loops in several outputs, except for the one shown in Figure 2C. Similarly, those of the 3 UTR were mostly inconsistent but a stem-loop predicted near the 3 -end was the most probable ( Figure 2C). It is known that some members of Tymoviridae, such as turnip yellow mosaic virus (TYMV), have tRNA-like structures (TLS) in their 3 -UTR that play an important role in viral replication [63,64]. Given that TLSs are difficult to predict (e.g., the TYMV-TRL (6234-6318 nt, NC_004063) was not detected by tRNAscan-SE software [65]), more extensive base-pairing analysis of the FbLFV1 3 -end should be conducted before drawing any definitive conclusions.

Molecular Phylogenetic Analysis of FbLFV1
Amino acid sequences of whole FbLFV1 replicase and representative mycoviruses in the order Tymovirales, including taxonomically unassigned members, were aligned and the Met, Hel and RdRp domains were extracted (Table 1 and Figure 3). These three domains were found to be well conserved in the deduced protein sequence of FbLFV1 replicase. The evolutionarily conserved relationship of FbLFV1 to the Tymovirales members was characterised by maximum likelihood (ML) phylogenetic analyses based on the alignments of each of the three domains ( Figure 4). As expected from BLAST analyses, the FbLFV1 Met Hel and RdRp domains were distantly associated with those of viruses in the order Tymovirales, with relatively close relationship to a gammaflexivirus ( Figure 4A-C). The concatenate sequence of these domains finally revealed that FbLFV1 was placed independently from the clades of established families in the order Tymovirales, although the branching support value was not sufficiently high in the ML tree ( Figure 4D). Based on these phylogenetic analyses, FbLFV1 is expected to belong to a novel virus species "Fusarium boothii large flexivirus 1" that may form an independent genus, or even a new virus family, in the order Tymovirales.   Table 1.

Analysis of Viral Sequences: Three Mitoviruses
The full-length sequences of the 3 kbp-dsRNAs in F. boothii strains Ep-BL13, Ep-BL14 and Ep-N28 were obtained by cDNA library sequencing and 3 RLM-RACE analyses (accession Nos. LC425112, LC425113 and LC425114, respectively). The viral dsRNAs accumulated in strains Ep-BL13 and Ep-N28 were 2802 bp long and that in strain Ep-BL14 was 2801 bp long. Each positive sense strand sequence possessed a single ORF potentially encoding RdRp when using mitochondrial codon usage ( Figure 5A). The length of the ORF and 3 -UTRs was the same among three sequences, being 2472 nt (823 aa) and 136 nt, respectively ( Figure S2). A few nucleotide deletions at different positions in the 5 -UTRs were found in the three viral sequences; nevertheless, the 5 -UTR of Ep-BL13 and Ep-N28 were of the same length (194 nt), whereas that of strain Ep-BL14 was 1 nt shorter (193 nt; Figure S2). The nucleotide similarities between these three viruses were 98% in any combinations. A BLASTP search found that the highest aa sequence identity of the RdRp coded by three viral RNAs of Ep-BL13, Ep-BL14 and Ep-N28 were 41% (E-value = 1 × e −158 ), 40% (E-values = 8× e −158 ) and 41% (E-value = 2× e −159 ), respectively, to the RdRp sequence of a tentative mitovirus of uncertain origin, soybean leaf-associated mitovirus 5. The highest aa identity with a definitive mitovirus species was 30% to Ophiostoma novo-ulmi mitovirus 3a (OnuMV3a). Thus, these 3 kbp-dsRNAs are apparently replication intermediates of new mitovirus strains that are closely related to each other, and these viruses were tentatively termed Fusarium boothii mitovirus 1 (FbMV1).
Some mitoviruses form a long pan-handle structure between the 5 and 3 termini [66]. In the terminal regions of FbMV1 genome, tetra nucleotides at extreme termini 5 -GGGG and CCCC-3 may have potential to form base pairs. Another common characteristic of mitovirus genomes is the presence of stem-loop structures at near the terminal regions. Both the 5 and 3 UTRs of the FbMV1 genome clearly folded into two potential stem-loop structures, as shown in Figure 5B. None of the sequences was predicted to form pseudoknot. and Ep-N28 were 2802 bp long and that in strain Ep-BL14 was 2801 bp long. Each positive sense strand sequence possessed a single ORF potentially encoding RdRp when using mitochondrial codon usage ( Figure 5A). The length of the ORF and 3′-UTRs was the same among three sequences, being 2472 nt (823 aa) and 136 nt, respectively ( Figure S2). A few nucleotide deletions at different positions in the 5′-UTRs were found in the three viral sequences; nevertheless, the 5′-UTR of Ep-BL13 and Ep-N28 were of the same length (194 nt), whereas that of strain Ep-BL14 was 1 nt shorter (193 nt; Figure S2). The nucleotide similarities between these three viruses were 98% in any combinations. A BLASTP search found that the highest aa sequence identity of the RdRp coded by three viral RNAs of Ep-BL13, Ep-BL14 and Ep-N28 were 41% (E-value = 1 × e −158 ), 40% (E-values = 8× e −158 ) and 41% (E-value = 2× e −159 ), respectively, to the RdRp sequence of a tentative mitovirus of uncertain origin, soybean leaf-associated mitovirus 5. The highest aa identity with a definitive mitovirus species was 30% to Ophiostoma novo-ulmi mitovirus 3a (OnuMV3a). Thus, these 3 kbp-dsRNAs are apparently replication intermediates of new mitovirus strains that are closely related to each other, and these viruses were tentatively termed Fusarium boothii mitovirus 1 (FbMV1). Some mitoviruses form a long pan-handle structure between the 5′ and 3′ termini [66]. In the terminal regions of FbMV1 genome, tetra nucleotides at extreme termini 5′-GGGG and CCCC-3′ may have potential to form base pairs. Another common characteristic of mitovirus genomes is the presence of stem-loop structures at near the terminal regions. Both the 5′ and 3′ UTRs of the FbMV1 genome clearly folded into two potential stem-loop structures, as shown in Figure 5B. None of the sequences was predicted to form pseudoknot.

Molecular Phylogenetic Analysis of Mitoviral RdRps
Phylogenetic analysis of the RdRps of the three new mitoviruses and some representative mitoviruses and narnaviruses (members of the family Narnaviridae) was conducted using PhyML ( Figure 6 and Table S1). According to Xu et al. (2015) [34], mitoviruses are divided into three major clades I, II and III, and the new mitoviruses were included in clade II accommodating OnuMV3a. Mitoviruses found from Fusarium species have been mostly grouped into clade III, however, FbMV1 together with recently reported Fusarium poae mitovirus 3 and 4 (FpMV3 and FpMV4 in Figure 6) [25] exemplified a broad diversity of Fusarium mitoviruses. Taken together, these data suggest that the three viruses characterised in this study are strains of a new virus species Fusarium boothii

Molecular Phylogenetic Analysis of Mitoviral RdRps
Phylogenetic analysis of the RdRps of the three new mitoviruses and some representative mitoviruses and narnaviruses (members of the family Narnaviridae) was conducted using PhyML ( Figure 6 and Table S1). According to Xu et al. (2015) [34], mitoviruses are divided into three major clades I, II and III, and the new mitoviruses were included in clade II accommodating OnuMV3a. Mitoviruses found from Fusarium species have been mostly grouped into clade III, however, FbMV1 together with recently reported Fusarium poae mitovirus 3 and 4 (FpMV3 and FpMV4 in Figure 6) [25] exemplified a broad diversity of Fusarium mitoviruses. Taken together, these data suggest that the three viruses characterised in this study are strains of a new virus species Fusarium boothii mitovirus 1 in the genus Mitovirus in the family Narnaviridae. This is consistent with the species demarcation criteria within the genus Mitovirus (more than 90% sequence identity with RdRp refers strains) in the 9th report of ICTV [29].
Viruses 2018, 10, x FOR PEER REVIEW 13 of 19 mitovirus 1 in the genus Mitovirus in the family Narnaviridae. This is consistent with the species demarcation criteria within the genus Mitovirus (more than 90% sequence identity with RdRp refers strains) in the 9th report of ICTV [29].  Table S1. Red letters represent mitoviruses characterized in this study.

Discussion
Virus hunting from fungal hosts has been widely conducted, and recent advances in sequencing technologies has accelerated the discovery of novel types of mycoviruses [1,17,25,43,67]. The viruslike sequences mining from NGS-assisted metadata are highly useful in understanding the diversity and evolution of viruses although there is an argument as to what kind of information can validate the presence of "actual viruses" and their classification, even within ICTV [68]. Indeed, recent examples of virus characterization in Rosellinia necatrix and S. sclerotiorum revealed identification of novel mycoviruses including taxonomically unique individuals, and demonstrated effectiveness of NGS [69,70]. Hence, the accumulation of more virome data from African biological materials will facilitate our understanding of mycovirus diversity and may lead to discovery of unexpectedly unique populations of mycoviruses, as is the case for FbLFV1. There are many barriers in the establishment of effective virocontrol of herbaceous annual plant diseases, including FHB, as discussed by Pearson et al. [71]. Thus, it will be important to screen for virocontrol agents which have much better properties such as higher transmissibility and broader cross-species specificity.
Characterization of mycoviruses has significantly contributed to the taxonomy of viruses including the order Tymovirales. Those include creation of the genus Mycoflexivirus and the family Gammaflexiviridae (BVF); creation of the genus Deltaflexivirus and the family Deltaflexiviridae (SsDFV1, SsDFV2, FgDFV1 and soybean leaf-associated mycoflexivirus 1); creation of the genera Figure 6. Phylogenetic analyses of FbMV1. A phylogenetic tree was constructed based on the alignment of RdRp domain sequences of narnaviruses, mitoviruses and FbMV1 strains. The midpoint-rooted ML tree is shown. Dot at the nodes represents reliable aLRT values over 0.9 for branch supporting. The best model of the ML tree constructs is indicated at the upper left. Clade clustering is referred to a previous report by Xu et al. (2015) [34]. Blue arrowheads indicate mitoviruses isolated from the Fusarium species. Abbreviations of viral names and accession numbers are listed in Table S1. Red letters represent mitoviruses characterized in this study.

Discussion
Virus hunting from fungal hosts has been widely conducted, and recent advances in sequencing technologies has accelerated the discovery of novel types of mycoviruses [1,17,25,43,67]. The virus-like sequences mining from NGS-assisted metadata are highly useful in understanding the diversity and evolution of viruses although there is an argument as to what kind of information can validate the presence of "actual viruses" and their classification, even within ICTV [68]. Indeed, recent examples of virus characterization in Rosellinia necatrix and S. sclerotiorum revealed identification of novel mycoviruses including taxonomically unique individuals, and demonstrated effectiveness of NGS [69,70]. Hence, the accumulation of more virome data from African biological materials will facilitate our understanding of mycovirus diversity and may lead to discovery of unexpectedly unique populations of mycoviruses, as is the case for FbLFV1. There are many barriers in the establishment of effective virocontrol of herbaceous annual plant diseases, including FHB, as discussed by Pearson et al. [71]. Thus, it will be important to screen for virocontrol agents which have much better properties such as higher transmissibility and broader cross-species specificity.
Characterization of mycoviruses has significantly contributed to the taxonomy of viruses including the order Tymovirales. Those include creation of the genus Mycoflexivirus and the family Gammaflexiviridae (BVF); creation of the genus Deltaflexivirus and the family Deltaflexiviridae (SsDFV1, SsDFV2, FgDFV1 and soybean leaf-associated mycoflexivirus 1); creation of the genera Sclerodarnavirus and Botrexvirus under the family Alphaflexiviridae (SsDARV and BVX, respectively); and proposal of new genus Mycotymovirus under the family Tymoviridae (FgMTV1) (Table 1; Figure 7) [20,27,[39][40][41][42]67]. This study additionally reports a novel member of the order Tymovirales that has a potential to further expand taxonomic diversity in the order.  Figure 7) [20,27,[39][40][41][42]67]. This study additionally reports a novel member of the order Tymovirales that has a potential to further expand taxonomic diversity in the order.  Table 1 and their genomic structures are illustrated. Viral families, subfamilies and genera are shown in bold letters, and 'new?' represents a potential new taxon. Colour differentiated symbols are as indicated. The arrow on BVF replicase indicates potential readthrough translation. The blue triangle points to a known proteolytic cleavage site. Gray triangles at 3′-end of FbLFV1 and DOSV genomes indicate uncommon form that is not polyadenylated or tRNA-like structured.  Table 1 and their genomic structures are illustrated. Viral families, subfamilies and genera are shown in bold letters, and 'new?' represents a potential new taxon. Colour differentiated symbols are as indicated. The arrow on BVF replicase indicates potential readthrough translation. The blue triangle points to a known proteolytic cleavage site. Gray triangles at 3 -end of FbLFV1 and DOSV genomes indicate uncommon form that is not polyadenylated or tRNA-like structured.
FbLFV1 showed similar characteristic to and distinct features from present five families in the order as follows. The careful assessment of phylogenetic relationship between FbLFV1 and other Tymovirales members revealed its close-yet-distinct correlations to the gammaflexivirus BVF and deltaflexiviruses (Figures 3 and 4). The presence of PHA03247 domain in these viruses further supported their evolutionally close association. However, the genome length and organization of FbLFV1 and BVF differed from each other (FbLFV1, 12,579 nt with single ORF vs. BVF, 6827 nt + poly-A tail with two ORFs) [39]. The absence of CP or CP domain of trichovirus/tymoviruses (see those of apple stem grooving virus, Diuris virus A and maize rayado fino marafivirus in Figure 7) in the FbLFV1 large polypeptide is highly possible and this feature is similar to deltaflexiviruses, an exceptional alphaflexivirus SsDRV and a tymo-like virus FgMTV1, highlighting the uniqueness of capsid-less nature of some mycoviruses, which is often observed (Figure 7) [20,27,38,41,42,45,67]. All Tymovirales viruses possess poly-A tail at the 3 end with exceptions of those of the genus Tymovirus in the family Tymoviridae, which have TLSs, and of the genus Platypuvirus in the family Alphaflexiviridae, which have no obvious RNA structures (Figure 7) [72]. The 3 end of FbLFV1 genome does has neither a poly-A tail nor TLS, further discriminating FbLFV1 from established genera and families in the order Tymovirales (Figures 2 and 7). In contrast, 5 ends of Tymovirales virus genomes poorly conserve characteristics, except for the cap structure, across the families. In the genus Tymovirus, 5 UTRs generally carry one to four protonatable hairpins nearby initiation codons that are probably required for dicistronic expression of overlapping ORFs (Figure 7, Tymovirus) [73]. The stem-loop structure predicted in 5 UTR of FbLFV1 was placed at 81 nt upstream of start codon ( Figure 2C) and there is no potential overlapping ORF, suggesting a distinct role of the structure in multiplication process of FbLFV1.
The genome length of the members in the order Tymovirales ranges from about 5.9 to 9.0 kb [59] whereas the recently reported unassigned member, a Rhizoctonia virus RsFV1, has the largest genome (10,621 nt + poly-A tail) [38]. The discovery of FbFLV1 increases the maximum genome size in the order by about 2 kb. RsFV1 and FbFLV1 share a similar genomic structure: the large RNA genome of over 10 kb; the single large ORF encoding replicase; the presence of a functionally unknown region at N-terminus in the replicase; and a potential of capsid-less nature (Figure 7). The N-terminal regions contain a PHA03247 protein motif that is similar to a portion of the UL36 large tegument protein found in herpesviruses that has also been found from other mycoviruses BVF and SsDRV (Figure 7) [74]. Despite the above-mentioned similarity, RsFV1 is evolutionally closer to deltaflexivirus, whereas FbLFV1 is phylogenetically independent. In addition, a taxonomically floating virus, OMSV, appeared to be monophyletic, despite its apparently tymovirus (Tymoviridae)-like characteristics with respect to genome organisation and spherical virion structure (Figures 4 and 7) [44]. Thus, FbLFV1 and OMSV will become key viruses when taxonomic reconsideration within the order Tymovirales is required.
The strains of FbMV1 were found in three independent F. boothii isolates, and TcMV exhibited a high similarity to FbMV1 (35%). It has been reported that the TcMV RdRp has an AroE domain, which is one of the domains on AROM protein encoded by Saccharomyces cerevisiae [75]. AROM is a pentafunctional polypeptide that catalyses a sequence of five reactions in prechorismate polyaromatic amino acid biosynthesis, which is the former half of shikimate pathway. Although the function of the AroE-like domain in TcMV RdRp is unclear, similar sequences have been found in phylogenetically close relatives such as CpMV1 and OnuMV3a, as well as FbMV1 ( Figure 5 and Figure  S3), further supporting the involvement of FbMV1 in the mitovirus group of clade II. TcMV was identified in a basidiomycete fungus, Thanatephorus cucumeris (anamorph Rhizoctonia solani), which is phylogenetically very distant from ascomycetous Fusarium species. These points support an earlier hypothesis by Deng et al. [76] that mitoviruses have evolved not only through co-evolution with the viruses strictly confined within their host fungi and close relatives, but also through host-switching and adaptation to phylogenetically distant hosts via unknown routes of transmissions.
Overall, this study sequenced three isolates of a new mitovirus FbMV1 harboured in three independent F. boothii strains, as well as a novel tymo-like virus, FbLFV1, with the largest genome size thus far described in the order Tymovirales. Detailed analyses of gene expression, potential polyprotein processing, particle formation and pathogenicity of FbLFV1 are the next challenges to understand the properties of this new taxon in the order Tymovirales. Biological property of FbLFV1 may be able to be assessed by protoplast-fusion method as reported for FgV1-DK21 [77].
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4915/10/11/584/ s1, Table S1. List of viruses in the family Narnaviridae and related unassigned members used in the phylogenetic study shown in Figure 6. Figure S1. Phylogenetic analysis of partial TEF1-α gene sequences derived from Fusarium samples. Products from the genomic PCR targeting TEF1-α were subjected to direct sequencing and the resultant nucleotide sequences were aligned in MAFFT ver. 7 with publicly available sequences of various F. graminearum species complex members. The alignment was used for maximum likelihood phylogenetic analysis with the recommended model shown at upper left. Figure S2. Nucleotide sequence alignment of three FbMV1 strains Ep-BL13, Ep-BL14 and Ep-N28. The entire nucleotide sequences of FbMV1 strains aligned with MAFFT ver.7. Small letters, sequence of non-coding region; capital letters, replicase coding region. Figure S3. The putative AROM domains in mitovirus RdRps. The amino acid sequence of AroE domain of AROM protein encoded by S. cerevisiae, and corresponding sequences of FbMV1, TcMV and close relatives were aligned with MAFFT ver.7.