Virome Identification and Characterization of Fusarium sacchari and F. andiyazi: Causative Agents of Pokkah Boeng Disease in Sugarcane

Fusarium sacchari and Fusarium andiyazi are two devastating sugarcane pathogens that cause pokkah boeng disease (PBD) in China. RNA_Seq was conducted to identify mycoviruses in F. sacchari and F. andiyazi isolates collected from PBD symptom-showing sugarcane plants across China. Fifteen isolates with a normal, debilitated, or abnormal phenotype in colony morphology were screened out for the existence of dsRNA from 104 Fusarium isolates. By sequencing the mixed pool of dsRNA from these Fusarium isolates, a total of 26 contigs representing complete or partial genome sequences of ten mycoviruses and their strains were identified, including one virus belonging to Hypoviridae, two mitoviruses with seven strains belonging to Narnaviridae, one virus of Chrysoviridae, and one alphavirus-like virus. RT-PCR amplification with primers specific to individual mycoviruses revealed that mitoviruses were the most prevalent and the alphavirus-like virus and chrysovirus were the least prevalent. In terms of host preference, more mitoviruses were found in F. andiyazi than in F. sacchari. Fusarium sacchari hypovirus 1 with a 13.9 kb genome and a defective genome of 12.2 kb, shares 54% identity at the amino acid level to the Wuhan insect virus 14, which is an unclassified hypovirus identified from insect meta-transcriptomics. The alphavirus-like virus, Fusarium sacchari alphavirus-like virus 1 (FsALV1), seemed to hold a distinct status amid fungal alphavirus-like viruses, with the highest identity of 27% at the amino acid level to Sclerotium rolfsii alphavirus-like virus 3 and 29% to a hepevirus, Ferret hepatitis E virus. While six of the seven mitoviruses shared 72–94% identities to known mitoviruses, Fusarium andiyazi mitovirus 2 was most similar to Alternaria brassicicola mitovirus with an identity of only 49% between the two viruses. Transmission of FsALV1 and Fusarium sacchari chrysovirus 1 (FsCV1) from F. sacharri to F. commune was observed and the characterization of the four-segment dsRNA chrysovirus was performed with aid of electron microscopy and analysis of the encapsidated RNAs. These findings provide insight into the diversity and spectrum of mycoviruses in PBD pathogens and should be useful for exploring agents to control the disease.

Fusarium sacchari and Fusarium andiyazi are two devastating sugarcane pathogens that cause pokkah boeng disease (PBD) in China. RNA_Seq was conducted to identify mycoviruses in F. sacchari and F. andiyazi isolates collected from PBD symptomshowing sugarcane plants across China. Fifteen isolates with a normal, debilitated, or abnormal phenotype in colony morphology were screened out for the existence of dsRNA from 104 Fusarium isolates. By sequencing the mixed pool of dsRNA from these Fusarium isolates, a total of 26 contigs representing complete or partial genome sequences of ten mycoviruses and their strains were identified, including one virus belonging to Hypoviridae, two mitoviruses with seven strains belonging to Narnaviridae, one virus of Chrysoviridae, and one alphavirus-like virus. RT-PCR amplification with primers specific to individual mycoviruses revealed that mitoviruses were the most prevalent and the alphavirus-like virus and chrysovirus were the least prevalent. In terms of host preference, more mitoviruses were found in F. andiyazi than in F. sacchari. Fusarium sacchari hypovirus 1 with a 13.9 kb genome and a defective genome of 12.2 kb, shares 54% identity at the amino acid level to the Wuhan insect virus 14, which is an unclassified hypovirus identified from insect meta-transcriptomics. The alphaviruslike virus, Fusarium sacchari alphavirus-like virus 1 (FsALV1), seemed to hold a distinct status amid fungal alphavirus-like viruses, with the highest identity of 27% at the amino acid level to Sclerotium rolfsii alphavirus-like virus 3 and 29% to a hepevirus, Ferret hepatitis E virus. While six of the seven mitoviruses shared 72-94% identities to known mitoviruses, Fusarium andiyazi mitovirus 2 was most similar to Alternaria brassicicola mitovirus with an identity of only 49% between the two viruses. Transmission of FsALV1 and Fusarium sacchari chrysovirus 1 (FsCV1) from F. sacharri to F. commune

INTRODUCTION
Pokkah boeng disease (PBD) is an airborne fungal disease caused by Fusarium fujikuroi species complex (FFSC) and is responsible for severe yield losses in susceptible varieties of sugarcane worldwide. The FFSC species F. sacchari is associated with both PBD and sugarcane wilting (Viswanathan et al., 2017) as well as fruit rot disease in bananas (Abd Murad et al., 2017), spear rot of oil palm (Suwandi et al., 2018), and even mycotic keratitis among sugarcane farmers (Bansal et al., 2016). The FFSC species F. andiyazi has been reported to cause wheat head blight (Wang et al., 2015), ear rot of maize (Bandara et al., 2017), and stalk rot of sorghum (Zhang et al., 2014). Recently, our team found F. sacchari and F. andiyazi strains with various phenotypes in Guangxi, Fujian, and Yunnan, which are the major sugarcane producing provinces in China (Wang J. et al., 2018;Meng et al., 2020). As one of the most important fungal diseases of sugarcane, PBD has long impacted sugarcane production and most sugarcane varieties grown in China are susceptible (Lin et al., 2014). Although the use of antifungal agents has been somewhat effective in controlling PBD, FFSC readily develops resistance to these treatments . Mycoviruses used as biocontrol agents could be an alternative approach to reduce the economic impact of F. sacchari and F. andiyazi on sugarcane crops.
RNA sequencing techniques have been used for the efficient discovery of novel viruses in fungi, including Benyviridae, Ophioviridae, and Virgaviridae (Marzano et al., 2016), as well as 10 viruses in five S. sclerotiorum strains (Khalifa et al., 2016), 17 mycoviruses in a F. poae strain (Osaki et al., 2016), and a large number of mycoviruses in Rhizoctonia solani AG2-2 LP isolates (Picarelli et al., 2019). In this study, we identified and characterized mycoviruses in the isolates of F. sacchari and F. andiyazi through RNA_Seq analysis and RT-PCR amplification.

Metatranscriptomic Identification of Mycoviruses in F. sacchari and F. andiyazi
The gel electrophoresis of dsRNA revealed that nine F. andiyazi and six F. sacchari from the 104 Fusarium isolates (42 F. sacchari, 41 F. andiyazi and 21 F. proliferatum) tested positive for dsRNA presence (Supplementary Figure S1). These dsRNAs were used for complete genome sequencing. After removal of low-quality and host reads, 42,947,006 and 27,056,546 reads from 67,969,464 and 79,658,766 clean reads (paired-end) were de novo assembled to produce 35,318 and 41,638 contigs for F. sacchari and F. andiyazi, respectively. Twenty-six contigs, representing partial or complete mycovirus genome segments, were obtained by BLAST alignment against a GenBank nt database ( Table 1). RT-PCR amplification further confirmed that these putative viruses indeed existed in the fungal isolates (Figure 1). Of the ten putative mycoviruses, nine were predicted to have (+) ssRNA genomes and the remaining one had a dsRNA genome. In the nine (+) ssRNA viruses isolated from the 15 fungal strains, seven were mitoviruses (2,145-2,463 nt), one (13,969 nt) was related to a hypovirus, and another (7,685 nt) was an alphavirus-like virus. The dsRNA virus had four segments (3,518, 2,796, 2,779, and 2,569 nt, respectively) and was most similar to chrysovirus in Chrysoviridae.

One Novel Virus in the Family Hypoviridae
A predicted hypovirus and its defective RNA were identified for F. sacchari FJ-FZ06 and GX-FS01. Contig 17922_seq1 and contig 10311_seq1 overlapped. Contig 17922_seq1 had 13,969 nt and contained one complete ORF that was predicted to encode a putative polyprotein of 4,257 aa. Based on BLASTp analysis, the viral-encoded polyprotein had the highest identity (54.1%, coverage 86%, E-value = 0) with the polyprotein encoded by Wuhan insect virus 14 (WIV14), a hypo-like virus identified from insect meta-transcriptomics (Shi et al., 2016). Additionally, this viral protein has identity (29-48%, coverage 45-81%, E-value = 0) to those of Alternaria alternata hypovirus 1 (AaHV1), CHV1, CHV2, and Fusarium graminearum hypovirus 1 (FgHV1) Hillman et al., 1994;Wang et al., 2013;Shi et al., 2016;Li et al., 2019; Supplementary Figure S2). The viral-encoded polyprotein was searched against the conserved domain in NCBI and exhibited three conserved domains, Peptidase_C7 (pfam01830), DUF3525 (pfam12039), and DEAH_box_HrpB (TIGR01970) with low or moderate E-values, which is similar to FgHV1. WIV14 and AaHV1 proteins (with the highest coverage and identity) also contained DUF3525 and DEAH_box_HrpB but lacked the Peptidase_C7 domain. The polyproteins encoded by ORF B of CHV1 and CHV2 had the Peptidase_C8 domain (Figure 2A). In contrast to the ORF B of CHV1 and CHV2, the RdRp domain was not detected in our strains. However, when the viral-encoded and WIV14 polyproteins were aligned with polyproteins of CHV1 and CHV2, nine RdRp core motifs (Ia-VIII) were identified in the region between the DUF3525 and helicase domains (including DEAH_box_HrpB and DEXDc) (Koonin et al., 1991;Supplementary Figure S2A). Based on this region, the viral sequence was grouped in a clade with WIV14, FgHV1, MpHV1 (Macrophomina phaseolina hypovirus 1), CHV1, and CHV2 from the phylogenetic tree of the members of family Hypoviridae ( Figure 2B; Yaegashi et al., 2012). Owing to its sequence relatedness to hypo-and hypo-like viruses, this newly identified virus is designated as Fusarium sacchari hypovirus 1 (FsHV1). Contig 17922_seq2 had 12,289 nt and contained one complete ORF encoding a putative polyprotein of 3,697 aa, which had a 54% sequence identity to WIV14. These two protein sequences were identical, except that the longer sequence (contig 17922_seq1) had an additional 560 aa at position 1,108. Due to the emergence of defective viral RNA, as was previously observed from other hypo-or hypo-like viruses , the shorter sequence (contig 17922_seq2) could be a defective RNA that has an internal genome deletion. We named contig 17922_seq2 Fusarium sacchari hypovirus 1-Defective RNA (FsHV1-D RNA) ( Table 1). FsHV1 was identified in F. sacchari strains FJ-FZ06, GX-FS01 and GX-FS03, whereas its D RNA was only detected in GX-FS01.
When the N-terminal region (145-240 aa region) of the FsHV1 putative polyprotein was aligned with papain-like cysteine protease domain regions of polyproteins from other hypo-and hypo-like viruses, the presence of conserved three cysteine protease core residues (cysteine, histidine, and glycine in Supplementary Figure S2B) indicated that FsHV1 encoded polyproteins processed by cysteine protease (Koonin et al., 1991). FsHV1 in the phylogenetic tree clustered with CHV1, CHV2, WIV14, AaHV1, and FgHV1 ( Figure 2B). This group has a conserved SDD tripeptide in RdRP motif VI (see Figure 2B), except for FgHV1, which is GDD.

One Novel and Six Previously Reported Mitoviruses in the Family Narnaviridae
From the RNAseq assembly, we identified 18 sequences encoding for proteins that showed high similarity with viruses from the Narna-levi clade (Shi et al., 2016), but several of them encoded the uncomplete ORFs of Mitovir_RNA_pol (Mitovirus RNAdependent RNA polymerase; pfam05919). To identify different isolates, we assembled and aligned all of the sequences and designed primers specific for conserved regions, including the complete ORFs. A total of 13 contigs originated from one F. sacchari isolate and five F. andiyazi isolates were assembled into seven putative mitoviruses ( Table 1). These putative mitovirus sequences from F. sacchari and F. andiyazi were 2.1 to 2.5 kb long, respectively, and encoded the complete Mitovir_RNA_pol domain: two for 708 aa protein, one for 709 aa protein, one for 724 aa protein, one for 727 aa protein, and one for 741 aa protein, suggesting that there might be multiple mitovirus isolates represented by these sequences. The RdRp catalytic motif GDD was searched against the conserved domain database of NCBI (Supplementary Figure S4). BLASTp showed that the top hit was to the Fusarium circinatum mitovirus 2-1 (FcMV2-1) (two contigs, 93 and 94% identity; E-values, 0.0); for the other four contigs, the top hit was the same as FcMV2-1 with lower identity (72∼73% identity; E-values, 0.0); and for the remaining contig, the top hit was to Alternaria brassicicola mitovirus (AbMV) (49% identity; E-value, 0.0). These viral sequences matched the typical pattern for fungal mitoviruses, in which 83% (10/12) of their Trp was encoded by UGA (Supplementary Figure S4). We subsequently included viral sequences encoding the Mitovir_RNA_pol domain in the phylogenetic analysis (Figure 4). All seven sequences clustered in one sub-clade. Six sequences and FcMV2-1 shared 72-94% identity, which were named Fusarium andiyazi mitovirus 1. The remaining sequence shared 49% identity of NoMV2 was named Fusarium andiyazi mitovirus 2 (FaMV2).

One Strain of a Previously Reported dsRNA Virus in the Family Chrysoviridae
Chrysoviridae is a family of dsRNA viruses with genomes ranging from 11.5 to 12.8 kb that typically include four encapsidated segments at sizes from 2.5 kb to 3.6 kb (Kotta-Loizou et al., 2020). A chrysovirus was detected in F. sacchari FJ-FZ04. Four contigs were associated with Fusarium oxysporum f. sp. dianthi mycovirus-1 (FodV1) ( Table 1). Contig 2971_seq1 having 3,518 nt contained a complete ORF encoding a putative protein of 1,139 aa that had the highest similarity to the FodV1 RdRp (89% identity). Contig 18164_seq1 consisted of 2,796 nt and contained a complete ORF encoding a putative 878 aa protein, which was similar to the hypothetical protein P2 of FodV1 (87% identity). Contig 17958_seq1 consisted of 2,779 nt and contained a complete ORF encoding a putative 852 aa protein that had the highest similarity to the FodV1 coat protein (91% identity; Table 1). Contig 3153_seq1 consisted of 2,569 nt and contained one complete ORF predicted to encode an 830 aa protein. Blastp analysis showed that this putative protein had sequence similarity to the hypothetical FodV1 protein P4 (92% identity). These four contigs represented a new FodV1 isolate from F. sacchari FJ-FZ04. Therefore, we named this sequence Fusarium sacchari chrysovirus 1 (FsCV1), which was closely related to FodV1 based on the phylogenetic tree constructed from the conserved RdRp domains of FodV1 and other selected chrysoviruses using a maximum-likelihood method (Figure 5). FsCV1 shared a similar genomic organization and size as FodV1.
Chrysoviridae family members could be divided into two distinct clusters from the phylogenetic tree (Zhai et al., 2018). Cluster I contained members of the genus Chrysovirus and 3-segmented chrysovirus-related unclassified viruses. FsCV1 grouped with FodV1 to form a separated evolutionary clade in cluster II, which  Supplementary Table S3. FsHV1 discovered in this work is marked with a red star.  Supplementary Table S3. FsALV1 discovered in this work is marked with a red triangle.
FIGURE 4 | Phylogenetic analysis of Mitoviruses. 91 sequences have been used to produce an alignment starting from viruses belonging to Narnaviruses, Ourmiaviruses and Ourmia-like viruses as outgroup; the phylogenetic tree was built using the neighbor-joining method, the best choice for a distance calculated with a Poisson model and uniform rates between sites using the program MEGA 6.0. Ultrafast bootstrap analysis was performed with 1,000 replicates. Viruses discovered in this work are marked with purple diamonds. Previously identified mitoviruses from Ascomycota hosts are labeled with red triangles. Other mitoviruses from other phyla of fungal hosts are labeled: Basidiomycota, purple circle; Glomeromycotina, blue triangle; Zoopagomycota, brown square; plants, green diamond. The names (full and abbreviated) of selected viruses as well as GenBank accession numbers are listed in Supplementary Table S3.  Supplementary Table S3. FsCV1 discovered in this work is marked with a red triangle.
contained chrysovirus-related, unclassified viruses with four to seven genomic segments (Figure 5).

Horizontal Transmission of FsALV1 and FsCV1 via Hyphal Fusion
The strain F. sacchari FJ-FZ04 was coinfected with FsALV1 and FsCV1, which were confirmed by RT-PCR detection (Figure 1). Virus particles of FsCV1 were purified from the mycelia of FJ-FZ04 using CsCl gradient centrifugation. Agarose gel electrophoresis of the nucleic acids extracted from the 1.30 -1.40 g cm −3 CsCl gradient fractions showed dsRNA bands of FsCV1 (Supplementary Figure S5B). The purified virus particles were isometric and approximately 50 nm in diameter as observed by transmission electron microscopy (Supplementary Figure S5A). RT-PCR amplification of dsRNA extracted from purified virions identified four dsRNA segments (Supplementary Figure S5C). Transmission of the virus from the F. sacchari strain FJ-FZ04 which harbored FsCV1 and FsALV1 to the virus-free recipient F. commune strain GX4-46, pathogen of sugarcane root rot (Wang J. et al., 2018), were conducted by pairing both strains on PDA plates for a week. After three consecutive subcultures of the recipient strain (designated as F. commune GX4-4-V), RT-PCR was performed with primers specific to the FsALV1 and FsCV1. PCR products of the expected sizes were amplified, demonstrating that both FsALV1 and FsCV1 were transmitted from FJ-FZ04 to GX4-46 via anastomosis. To exclude the possibility of contamination from the donor strain, the newly obtained virus-harboring F. commune GX4-4-V was subjected to re-identification by sequencing the TEF1-α gene which was amplified with primer pairs EF1/EF2. The result showed that TEF1-α from F. commune GX4-4-V was identical to that of the F. commune GX4-46 and shared 88.9% identity with the donor strain F. sacchari FJ-FZ04. Compared with the virus-free F. commune GX4-46, F. commune GX4-4-V grew slightly faster on PDA plate, but the pigmentation was slightly reduced (Figure 6).

DISCUSSION
A complex virome and its diverse viral segments were identified and characterized from 15 isolates of F. sacchari and F. andiyazi using high-throughput transcriptome sequencing. Some of these detected viruses were predicted to belong to the viral species within the families Hypoviridae, Narnaviridae, and Chrysoviridae, whereas one was an unclassified alphavirus-like (+) ssRNA virus. To the best of our knowledge, this is the first report of a comprehensive analysis on the viral diversity in the isolates of F. sacchari and F. andiyazi. Among six F. sacchari isolates, one hypovirus and one novel alphavirus-like virus as well as one characterized chrysovirus and one mitovirus were detected. Another six characterized mitoviruses and one novel mitovirus were present among nine F. andiyazi isolates.
The hypoviruses are known to be associated with hypovirulence and other symptoms, including slow growth, reduced conidiation, or toxin inhibition, in C. parasitica, S. sclerotiorum, and F. graminearum (Li et al., 2015;Marzano et al., 2016;Rigling and Prospero, 2018). The hypoviruses were clustered into two clades based on the RdRp and Hel domains of the polyprotein regions in the recent ICTV report (Suzuki et al., 2018). In this study, FsHV1 clustered with CHV1 and CHV2 in the "alphahypovirus" group. The Peptidase_C7 was found in the 5 terminus of the FsHV1 encoding polyprotein as well as the 5 terminus of ORFA of CHV1. The product of the Peptidase_C7 domain in CHV1 is p29 protein, which is a protease acting as a virus-encoded determinant that can alter fungal host phenotypes in CHV1 isolates (Xiong et al., 2019). The p29 protein was able to suppress both virus-induced and agroinfiltration-induced RNA silencing and systemic spread of silencing in GFP-expressing transgenic Nicotiana benthamiana line 16c plants (Segers et al., 2006). In addition, p29 has been shown to stimulate membrane and vesicle proliferation by directing p29 to the trans-Golgi network (TGN) vesicles through cofractionation with TGN membranes (Jacob-Wilk et al., 2006). ORF B was associated with virus transmission and host hypovirulence by affecting the transmission efficiency of conidiospores and viral accumulation (Lin et al., 2007). Although FJ-FZ06 caused lesions in sugarcane stalks, its virulence awaited certification after the construction of virus-free strains.
Both complete and defective viral genomes of FsHV1 were detected in GX-FS09. For defective FsHV1, no conserved domain was found in the absent region between genome coordinates 4,058 and 5,738. D-RNAs were also detected in CHV1, AaHV1, and other family members, such as Rosellinia necatrix partitivirus 2 (RnPV2). RnPV2 D-RNAs affect parental virus replication and mitigate viral symptoms in a Dicer-like 2 knockout mutant of C. parasitica (chestnut blight fungus), which is an artificial host for RnPV2 (Chiba et al., 2016). The discovery of D-RNAs in F. sacchari is very important, although further research is required to clarify whether FsHV1-S D-RNA affects viral replication and the fungal host phenotype. The availability of two fully assembled FsHV1 genomes allows for comparison with other members of the family Hypoviridae.
Mitoviruses have small plus-strand RNA genomes and are present in various species of fungi. RdRps from FaMV1-DH06 and -DZ were related to FcMV2-1 with 91-92% sequence identity, whereas FsMV1-LC04, FaMV1-SJ46, -BS38, and -GM64 were related to FcMV2-1 with 72-73% identity, suggesting that various FcMV2-1 strains were widely distributed in Yunnan, Guangxi, and Hainan Provinces in China. F. andiyazi isolates also appeared more frequently in samples from Yunnan. Another novel mitovirus, FaMV2, was also present in Yunnan Province. The observation that mitoviruses transmit vertically together with host propagation could provide some clues as to why Fusarium spp. was more divergent in Yunnan Province. Mitoviruses appeared to reduce virulence (hypovirulence) in various plant pathogens, including C. parasitica, Ophiostoma novo-ulmi, S. homoeocarpa, and S. sclerotiorum (Polashock and Hillman, 1994;Hong et al., 1998;Deng and Boland, 2006;Xu et al., 2015). The viral transmission and pathogenicity of mitoviruses from F. andiyazi and F. sacchari will need to be explored in more detail. Fusarium spp. mitoviruses mainly fell into two sub-clades of the phylogenetic tree, suggesting that their interspecific transmission might not occur as easily in nature as other fungi, such as Rhizoctonia solani and Sclerotinia sclerotiorum (Mu et al., 2017;Picarelli et al., 2019).
The F. sacchari strain FJ-FZ04 was found to be infected by two viruses, one chrysovirus and one unclassified alphaviruslike virus. FsCV1 shared the highest sequence identity with the previously reported virus FodV1, which is associated with virulence and other phenotypic traits of the plant pathogenic fungus F. oxysporum f. sp. dianthi. FsCV1 grouped with FodV1, FgV2, and FgV-ch9 in cluster II in the phylogenetic tree. These three mycoviruses were confirmed to be associated with phenotypic alterations, including hypovirulence. A high level of viral accumulation of FodV1 has been reported to reduce mycelial growth, conidiation, and virulence (Lemus-Minor et al., 2018). These results were comparable with the previous report for FgV-ch9, whereby higher accumulation in F. graminearum was linked to a reduction in conidiation and virulence (Darissa et al., 2011). Thus, it may be interesting to investigate whether FsCV1 can induce hypovirulence in F. sacchari. FsALV1 was also found in FJ-FZ04, but whether FJ-FZ04 hypovirulence is also associated with FsALV1 or not will require further studies.
In conclusion, various mycoviruses were detected in F. sacchari and F. andiyazi, the pathogenic fungi that causes pokkah boeng disease in sugarcane in China. Our results confirmed that mycovirus infection of F. sacchari and F. andiyazi was widespread in leading sugarcane producing regions and that some of these viruses might be associated with the hypovirulence of Fusarium spp.

MATERIALS AND METHODS
Growth of F. sacchari and F. andiyazi Isolates Fusarium sacchari and F. andiyazi isolates were recovered from the diseased leaves of sugarcane (see Supplementary Table S1) and grown on potato dextrose agar (PDA) medium at 28 • C. Isolates were kept on PDA plates at 4 • C for this study.

Total RNA Extraction and Purification
Each of the 15 isolates was cultured on a PDA plate for 5-7 days. Total RNA was extracted from the one-gram mycelial mass using a Trizol kit (Takara Bio, Inc., Japan) according to the manufacturer's instructions. Total RNA (1 µg) from each of the 15 isolates was mixed for RNA_Seq analysis.

RNA Sequencing
Sequencing was carried out by Beijing Novel Bioinformatics Co., Ltd., using an Illumina HiSeq 2500 instrument. Sequencing libraries were constructed from 15 mixed rRNA-depleted total RNA samples using a TruSeqTM RNA Sample Prep Kit (Illumina, RS-122-2001). After discarding low-quality reads, including paired-end reads less than 100 bp, reads with quality scores <20, and reads containing adapter sequences, clean reads were mapped to the genome of Fusarium spp. using bowtie2 v2.1.0. Reads that did not align concordantly were assembled by a metagenomic de novo assembly using Trinity (version: Trinity RNAseq-r2013-02-25). Large contigs with lengths ≥200 nt were obtained and subjected to local BLAST against a nucleotide database using BLASTn. A second alignment was performed against viral protein sequences using BLASTx, which was also used to remove false-positive hits against the nr database. To identify highly divergent viruses, contigs with the best match to viral protein or viral nucleotide sequences and E-values <1 × 10 −3 were retained.

Phylogenetic Analysis
The highly homologous viral nucleotide sequences or their deduced amino acid sequences of contigs were aligned against known viral nucleic acids and proteins in GenBank. The phylogenetic trees from the viruses and accession numbers of the viral gene(s) (Supplementary Table S3) were constructed using the Maximum likelihood method or Neighbor-Joining method with a bootstrap value of 1,000 replicates through MEGA 7.0 1 .

Purification of Virus Particles
Frozen mycelia (6 g) were homogenized in 0.1 M sodium phosphate buffer (pH 7.4) with a mixer at a frequency of 25 Hz for 1 min. The suspension was shaken for 30 min at 4 • C, mixed with 20% (v/v) of (1:1) chloroform:n-butanol, and centrifuged at 8,000 × g for 10 min. The chloroform:n-butanol step was repeated 2-3 times until the red pigment was removed from the supernatant. PEG6000 and NaCl were added to the supernatant to final concentrations of 8% (w/v) and 1% (w/v), respectively. After 1 h of incubation at 4 • C, the precipitate was collected by centrifugation (10,000 × g at 4 • C for 15 min), and the pellet was resuspended in 10 mL of 0.05 M sodium phosphate buffer (pH 7.0) at 4 • C for 3 h. The pellet was resuspended in 1 mL of 0.05 M sodium phosphate buffer (pH 7.0) and further centrifuged using CsCl density equilibrium centrifugation (25,000 rpm at 5 • C for 22 h). Virus-like particles were stained with 2% (w/v) 1 http://www.megasoftware.net/megamacBeta.php phosphotungstic acid and examined by transmission electron microscopy (HT-7700).

Hyphal Fusion
Virus infection via hyphal anastomosis was performed using the virus strain FJ-FZ04 as the donor and the virus-free Fusarium commune isolate GX4-46 as the recipient. Mycelial plugs of donor and recipient isolates were placed 2 cm apart on a 9 cm PDA plate and incubated at 28 • C for 7 days. Several plugs picked up from the border of strain GX4-46 in the region between the two colonies were sub-cultured three times on PDA plates.