Similar Characteristics of siRNAs of Plant Viruses Which Replicate in Plant and Fungal Hosts

Simple Summary RNA silencing in fungi was shown to confer antiviral defense against plant viruses. In this study, using high-throughput sequencing and bioinformatic analyses, we showed that small interfering RNAs (siRNAs) of cucumber mosaic virus and tobacco mosaic virus (TMV) which replicated in phytopathogenic fungi Rhizoctonia solani and Fusarium graminearum had similarities with viral siRNAs produced in plant hosts in regard to the size distributions, proportion of plus and minus senses, and nucleotide preference for the 5′ termini. Additionally, our results also determined that both F. graminearum DCL1 and DCL2 were involved in the production of TMV siRNAs. Thus, the fungal RNA silencing machineries have adaptive capabilities to recognize and process the genome of invading plant viruses. Abstract RNA silencing is a host innate antiviral mechanism which acts via the synthesis of viral-derived small interfering RNAs (vsiRNAs). We have previously reported the infection of phytopathogenic fungi by plant viruses such as cucumber mosaic virus (CMV) and tobacco mosaic virus (TMV). Furthermore, fungal RNA silencing was shown to suppress plant virus accumulation, but the characteristics of plant vsiRNAs associated with the antiviral response in this nonconventional host remain unknown. Using high-throughput sequencing, we characterized vsiRNA profiles in two plant RNA virus–fungal host pathosystems: CMV infection in phytopathogenic fungus Rhizoctonia solani and TMV infection in phytopathogenic fungus Fusarium graminearum. The relative abundances of CMV and TMV siRNAs in the respective fungal hosts were much lower than those in the respective experimental plant hosts, Nicotiana benthamiana and Nicotiana tabacum. However, CMV and TMV siRNAs in fungi had similar characteristics to those in plants, particularly in their size distributions, proportion of plus and minus senses, and nucleotide preference for the 5′ termini of vsiRNAs. The abundance of TMV siRNAs largely decreased in F. graminearum mutants with a deletion in either dicer-like 1 (dcl1) or dcl2 genes which encode key proteins for the production of siRNAs and antiviral responses. However, deletion of both dcl1 and dcl2 restored TMV siRNA accumulation in F. graminearum, indicating the production of dcl-independent siRNAs with no antiviral function in the absence of the dcl1 and dcl2 genes. Our results suggest that fungal RNA silencing recognizes and processes the invading plant RNA virus genome in a similar way as in plants.

shown to be capable of replicating in Colletotrichum acutatum and Phytophthora infestans (oomycete) [61,62]. TMV was also shown to replicate in F. graminearum by inoculation through the transfection of fungal protoplasts [63]. Moreover, C. parasitica, Valsa mali, F. graminearum, and the oomycete Phytoptora infestans can also host plant viroids, the subviral agents [64,65]. Importantly, our study previously discovered a natural infection of cucumber mosaic virus (CMV, genus Cucumovirus) in Rhizoctonia solani strains isolated from a potato plant [66]. Following this study, we recently also showed that by the screening of fungal strains isolated from virus-infected plants, various plant viruses were found to be commonly acquired by fungi [67]. These findings indicate that plant viruses can be transferred from plant to fungus during fungal colonization of the plant. TMV infection in C. acutatum induced gene silencing (virus-induced gene silencing) and is associated with the accumulation of vsiRNAs [61]. Deletion of both dcl1 and dcl2 genes in F. graminearum drastically elevated accumulations of TMV and hop stunt viroid [63,64]. Together, these observations indicate that fungal antiviral RNA silencing can recognize and target the invading plant viruses including viroids, but the characteristics of vsiRNAs associated with antiviral responses against plant virus infection in fungi have not been studied in detail.
In this study, we used deep sequencing analysis to characterize siRNAs derived from CMV and TMV that replicated in R. solani and F. graminearum, respectively. In addition, F. graminearum mutants defective in dcl genes were also included in the analyses. Our results revealed that the features of CMV and TMV siRNAs accumulated in fungi were similar to those accumulated in plants and confirmed the role of F. graminearum DCL1 and DCL2 in the biosynthesis of vsiRNAs with effective antiviral function.

Fungal Strains
R. solani strain Ra1 infected with CMV Rs isolates and CMV cured, and R. solani strain infected with CMV Fny isolate, have been described previously [66]. Fusarium graminearum (PH-1 strain) wild type, ∆dcl1, ∆dcl2, and ∆dcl1-∆dcl2 mutants infected with TMV have been described previously [63]. R. solani strain 80 (R. solani 80) was isolated from one of the potato tubers showing black scurf disease collected from potato-planting areas located in the middle and western parts of Inner Mongolia Province of China from the years 2008-2015 [66]. The presence of a virus in R. solani 80 was identified through high-throughput sequencing carried out in the previous study [66].

Plant Virus Inoculation
To inoculate plant viruses, CMV-and TMV-infected leaves were homogenized in 0.1 M phosphate buffer (pH 7.0) and rubbed onto carborundum-dusted leaves of N. benthamiana and N. tabacum plants, respectively. Noninoculated upper leaves were sampled seven days after inoculation.

Total RNA Extractions
Fungal total RNA was extracted from the mycelia of five-day-old PDB culture as described previously [63]. Briefly, fungal mycelia were homogenized in a buffer containing 100 mM Tris/HCl (pH 8.0), 200 mM NaCl, 4 mM EDTA, and 4% SDS, followed by phenol/chloroform extraction. After extraction, total RNA was treated with RQ1 DNase I to remove fungal DNA. Plant total RNA was extracted from leaves using TRIzol (Invitrogen, Waltham, MA, USA).

sRNA Sequencing
The cDNA libraries of sRNA were prepared using Truseq SRNA Sample Preparation Kit (Illumina, San Diego, CA, USA) according to the supplied protocol. The sRNA cDNA library was amplified with TruSeq PE Cluster Kit (Illumina, San Diego, CA, USA) and then used for sequencing on an Illumina HiSeq2500 at Hanyu Bio-Tech (Shanghai, China).

Bioinformatics Analysis
Adaptor sequences and poor-quality reads were removed by FASTX-toolkit (http: //hannonlab.cshl.edu/fastx_toolkit/, Version 0.0.14, accessed on 15 September 2021). Reads which were less than 18 nt or more than 30 nt were removed using the Cutadapt software (https://cutadapt.readthedocs.io/en/stable/, Version 4.1, accessed on 15 September 2021). The remaining reads with lengths of 18-30 nucleotides were mapped using Bowtie software (http://bowtie-bio.sourceforge.net, Version 1.3.1, accessed on 15 September 2021) to the genome of TMV (NC_001367.1) with 1 mismatch allowed, CMV isolate Rs (RNA1 (MG025947), RNA2 (MG025948), and RNA3 (MG025949)), CMV isolate Fny (RNA1 (D00356), RNA2 (NC_002035) and RNA3 (D10538)) and Endorna-like virus (OP763640) with 0 mismatches allowed. The number of vsiRNA reads was normalized to "reads per million" (RPM) according to the number of total reads of the corresponding sRNA library. Homemade Perl scripts and Excel were used to analyze the 5 -terminal nucleotide and the distribution of vsiRNA in the genome. The ggplot2 package in R (Version 4.1.1) was used to facilitate making the plot of siRNA mapping. Raw data of sRNA libraries have been deposited to the SRA database with the accession number PRJNA900007.

Characteristics of CMV-and Mycovirus-Derived siRNAs in R. solani
In a previous study, we discovered CMV infection in an R. solani strain (designated as Ra1 strain) isolated from a potato plant collected from a field in the Inner Mongolia Province of China [66]. This fungi-infecting CMV (referred to as Rs isolate) had a typical three-segmented RNA genome ( Figure 1A) and showed the closest identity (>98%) to East Asian CMV isolates belonging to the subgroup Ia. CMV-cured Ra1 strains were obtained through single spore isolation. In that study, we also artificially obtained a CMV-infected R. solani strain by transfection of protoplasts of a virus-free R. solani strain with CMV Fny isolate (subgroup I CMV isolate).
To characterize CMV siRNAs in R. solani, total RNAs were extracted from R. solani Ra1 and R. solani Ra1/CMV-cured (two biological replicate samples) and R. solani/CMV Fny and then subjected to high-throughput sequencing of sRNAs followed by bioinformatic analyses to investigate the sequence characteristics of vsiRNAs. Total RNA extracted from N. benthamiana infected with CMV Fny was also included in the high-throughput sequencing. In addition, to characterize mycovirus siRNAs in R. solani, high-throughput sequencing was also performed on total RNA extracted from an R. solani strain (R. solani 80) infected with a novel mycovirus related to endornaviruses (Supplementary File S1), tentatively named Rhizoctonia solani endorna-like virus 1 (RsEnLV1, Figure 1A). Sequence analysis yielded around 10-25 million sRNA reads ranging from 18 to 30 nucleotides in size for each sRNA library ( Table 1). Mapping of sRNA reads to the CMV genome revealed that high numbers of sRNAs were derived from CMV sequences in CMV-infected samples (R. solani Ra1 and R. solani/CMV Fny libraries) but no, or only very few, sRNAs were mapped to the CMV genome in R. solani Ra1/CMV-cured libraries ( Table 1). This result suggests that CMV infection in R. solani is associated with the accumulation of vsiRNAs. The proportion of CMV siRNAs was higher in the R. solani/CMV Fny library (0.07%) than in the R. solani Ra1 libraries (0.03% and 0.02%), but both were much lower than that in the N. benthamiana/CMV Fny library, which was around half (53.03%) of the total sRNA reads (Table 1 and Figure 1B). The proportion of positive-strand (+) CMV siRNAs was similarly higher than negative-strand (−) siRNAs in all libraries ( Figure 1B). Notably, much larger portions of CMV siRNAs in fungal samples were derived from RNA3 than, in order, RNA2 and RNA1. These differences were more pronounced in R. solani Ra1 than in the R. solani/CMV Fny library, whereas N. benthamiana/CMV Fny samples had similar proportions of RNA3-and RNA2-derived CMV siRNAs ( Figure 1C). It is still unclear whether the higher proportion of CMV RNA3-derived siRNAs in the fungal host is related to the higher accumulation of RNA3 in fungi or due to other reasons. High numbers of sRNA reads in the R. solani 80 libraries were also mapped to RsEnLV1 genome sequence, whereas a much lower number of sRNA reads in the R. solani 80/virus-cured libraries were mapped to RsEnLV1 genome sequence (Table 1 and Figure 1D), suggesting that RsEnLV1 infection in R. solani is associated with accumulation of vsiRNAs. libraries ( Figure 1B). Notably, much larger portions of CMV siRNAs in fungal samples were derived from RNA3 than, in order, RNA2 and RNA1. These differences were more pronounced in R. solani Ra1 than in the R. solani/CMV Fny library, whereas N. benthamiana/CMV Fny samples had similar proportions of RNA3-and RNA2-derived CMV siR-NAs ( Figure 1C). It is still unclear whether the higher proportion of CMV RNA3-derived siRNAs in the fungal host is related to the higher accumulation of RNA3 in fungi or due to other reasons. High numbers of sRNA reads in the R. solani 80 libraries were also mapped to RsEnLV1 genome sequence, whereas a much lower number of sRNA reads in the R. solani 80/virus-cured libraries were mapped to RsEnLV1 genome sequence (Table 1 and Figure 1D), suggesting that RsEnLV1 infection in R. solani is associated with accumulation of vsiRNAs.    Figure 2A). Differently, RsEnLV1 siRNAs in the R. solani 80 libraries had a broader size distribution; they were predominantly 21-24 nucleotides long, with peaks at 23 nucleotides ( Figure 2B). The 5′-terminal nucleotides of CMV siRNAs in the R. solani Ra1, R. solani/CMV Fny, and N. benthamiana/CMV Fny libraries had similar preferences; they were most biased toward C followed sequentially by U, A, and G ( Figure 2C). In the R. solani 80 libraries, the 5′-terminal nucleotide of RsIM-as EnV1 siRNAs was most frequently U ( Figure 2D). Thus, along with differences in size distribution, the 5′-terminal nucleotide preferences of siR-  The 5 -terminal nucleotides of CMV siRNAs in the R. solani Ra1, R. solani/CMV Fny, and N. benthamiana/CMV Fny libraries had similar preferences; they were most biased toward C followed sequentially by U, A, and G ( Figure 2C). In the R. solani 80 libraries, the 5 -terminal nucleotide of RsIM-as EnV1 siRNAs was most frequently U ( Figure 2D). Thus, along with differences in size distribution, the 5 -terminal nucleotide preferences of siRNAs between CMV and mycoviruses in R. solani are also different.
CMV 21 and 22 nt siRNAs were distributed along the (+) and (−) strands of the CMV RNA1, 2, and 3 genomes, with several prominent siRNA hotspots, most of which were located at the same genome positions across the R. solani Ra1 and R. solani/CMV Fny libraries, while the distribution and hotspots of CMV siRNAs from the N. benthamiana/CMV Fny libraries along the (+) and (−) strands of the genomes were much more dense, due to the high number of siRNAs ( Figure 3). Notably, in the fungal libraries, there was a prominent siRNA hotspot in the middle of the (+) strand of the CMV RNA3 genome, located at nucleotide position 950-971, which is the beginning of intercistronic region, but upstream of the initiation site of CMV CP subgenomic RNA synthesis [68]. Thus, the occurrence of this siRNA hotspot is not likely associated with CP subgenomic RNA accumulation. Similarly, RsEnLV1 siRNAs were also distributed along the (+) and (−) strands of virus genomes with several prominent siRNA hotspots (Supplementary Figure S1).

Characteristics of TMV siRNAs in Wild-Type and Dcl Mutants of F. graminearum.
In a previous study, we reported that F. graminearum is a compatible host for TMV replication [63]. TMV was introduced by protoplast transfections to F. graminearum strain PH-1 wild-type, single dcl knockout mutants (Δdcl1 and Δdcl2), and a double dcl knockout mutant (Δdcl1-Δdcl2) [63]. To characterize TMV siRNAs in F. graminearum, total RNAs were extracted from each TMV-infected wild-type and dcl knockout mutant strain, including the virus-free (VF) fungus, and then subjected to high-throughput sequencing of sRNAs. Additionally, total RNA samples extracted from N. tabacum infected with TMV

Characteristics of TMV siRNAs in Wild-Type and Dcl Mutants of F. graminearum
In a previous study, we reported that F. graminearum is a compatible host for TMV replication [63]. TMV was introduced by protoplast transfections to F. graminearum strain PH-1 wild-type, single dcl knockout mutants (∆dcl1 and ∆dcl2), and a double dcl knockout mutant (∆dcl1-∆dcl2) [63]. To characterize TMV siRNAs in F. graminearum, total RNAs were extracted from each TMV-infected wild-type and dcl knockout mutant strain, including the virus-free (VF) fungus, and then subjected to high-throughput sequencing of sRNAs. Additionally, total RNA samples extracted from N. tabacum infected with TMV were also included for sequence analysis. Two biological replicate samples were collected from each TMV-host combination and analyzed independently.
Sequence analysis yielded around 7-15 million sRNA reads ranging from 18 to 30 nucleotides in size for each sRNA library ( Table 2). Mapping of sRNA reads to the TMV genome revealed that no sRNA in the F. graminearum wild-type VF sRNA libraries was mapped to the TMV genome, while considerable numbers of sRNAs in the TMV-infected F. graminearum wild-type sRNA libraries (0.02% and 0.04%) were derived from TMV genome sequences, albeit far fewer than those mapped to the TMV genome in the TMV-infected N. tabacum sRNA libraries (27.33% and 10.74%) ( Table 2 and Figure 1E). Notably, there were far fewer TMV siRNAs in either F. graminearum single ∆dcl1 or ∆dcl2 mutant sRNA libraries than in the wild-type samples; surprisingly, however, TMV siRNAs were abundant in the double ∆dcl1-∆dcl2 mutant sRNA libraries, to an even greater extent than in wild-type sRNA libraries (Table 2 and Figure 1E). This result shows that deletion of either ∆dcl1 or ∆dcl2 drastically reduces TMV siRNA accumulation, but in the absence of both ∆dcl1 and ∆dcl2, TMV-derived sRNAs are produced through a DCL-independent pathway. TMV siRNAs in the F. graminearum wild-type and all ∆dcl mutant libraries were predominantly 21 and 22 nucleotides long, with peaks at 21 nucleotides ( Figure 4A), similar to the size distribution of those in TMV-infected N. tabacum libraries ( Figure 4B). The 5terminal nucleotide of TMV siRNAs in the F. graminearum wild-type and all ∆dcl mutant libraries as well as in TMV-infected N. tabacum libraries had a similar preference, with U the most frequent, followed sequentially by A, C, and G ( Figure 4C). Interestingly, TMV siRNAs between the F. graminearum wild-type and double ∆dcl1-∆dcl2 mutant libraries had very similar distribution profiles along (+) and (−) strands of the TMV genome, as the position of major and minor siRNA hotspots were parallel between these libraries ( Figure 5), while their distribution profiles were also relatively similar to those in the TMV-infected N. tabacum libraries, particularly the Nt-2 library, but less similar to those in the Nt-1 library, which had much more abundant TMV siRNAs than the Nt-2 library (Figures 1E and 5). Overall, these results showed that TMV siRNAs produced in F. graminearum have similar characteristics to those produced in plant hosts.

Discussion
The siRNAs and other classes of sRNAs play a central role in the RNA silencing pathway [1,5]. It is known that plants encode at least four DCL proteins [69]. In Arabidopsis thaliana, DCL4 and DCL2 are involved in antiviral defense against RNA viruses. DCL4 plays the major role in intracellular antiviral silencing and generates 21-nucleotide-long vsiRNAs, while DCL2 generates 22-nucleotide-long vsiRNAs and functions in intercellular (systemic) antiviral silencing. However, DCL2 can substitute for the antiviral functions of DCL4 when the latter is absent or repressed, although with lower efficiency [70][71][72][73][74][75]. DCL3 is involved in antiviral responses against DNA viruses and produces 24-nucleotidelong vsiRNAs [76,77]. Thus, RNA virus siRNAs in plants are predominantly 21 or 22 nucleotides in length, with peaks at 21 nucleotides. On the other hand, numerous studies have revealed that RNA mycovirus-derived sRNAs tend to have varied size profiles and

Discussion
The siRNAs and other classes of sRNAs play a central role in the RNA silencing pathway [1,5]. It is known that plants encode at least four DCL proteins [69]. In Arabidopsis thaliana, DCL4 and DCL2 are involved in antiviral defense against RNA viruses. DCL4 plays the major role in intracellular antiviral silencing and generates 21-nucleotide-long vsiRNAs, while DCL2 generates 22-nucleotide-long vsiRNAs and functions in intercellular (systemic) antiviral silencing. However, DCL2 can substitute for the antiviral functions of DCL4 when the latter is absent or repressed, although with lower efficiency [70][71][72][73][74][75]. DCL3 is involved in antiviral responses against DNA viruses and produces 24-nucleotide-long vsiRNAs [76,77]. Thus, RNA virus siRNAs in plants are predominantly 21 or 22 nucleotides in length, with peaks at 21 nucleotides. On the other hand, numerous studies have revealed that RNA mycovirus-derived sRNAs tend to have varied size profiles and broad size distributions among different viruses and fungal hosts. In F. graminearum, siRNAs derived from different mycoviruses also have broad size distributions, with a slight peak at 21 nucleotides [41,55]. Characterization of vsiRNA derived from taxonomically diverse mycoviruses infecting Botrytis cinerea showed that 20, 21, and 22 nucleotides are the predominant sizes, with the most abundant size class at 21 nucleotides, or 22 nucleotides for a certain mycovirus [53]. In Aspergillus fumigatus, the most abundant vsiRNA size class is 20, 21, or 20-23 nucleotides, depending on the virus [54]. Likewise, siRNAs of mitoviruses in Fusarium circinatum and C. parasitica have been found to have broad size distributions [46,58]. Together, these observations suggest that RNA genomes of various mycoviruses are processed differently by the fungal DCL to generate vsiRNAs.
Our results showed that the abundances of CMV and TMV siRNAs in the fungal hosts were much lower compared to those in the host plants ( Figure 1B,E). This is likely due to the lower CMV and TMV accumulation in fungi than in plants as observed in our previous studies [63,66]. It is intriguing that the characteristics of CMV and TMV siRNAs generated in the fungal hosts are distinct from the siRNAs of mycoviruses, but instead similar to the characteristics of CMV and TMV siRNAs generated in plant hosts. This may suggest that fungal DCLs have adaptive capabilities in the manner of recognizing and processing various types of incoming viruses. Although the antiviral role of DCLs of some ascomycetous fungi has been elucidated, the signatures of vsiRNA products of a particular fungal DCL are still unclear. In F. graminearum, DCL1 and DCL2 function redundantly in antiviral defense against TMV infection [63]. The results of our analysis on TMV siRNAs in F. graminearum single ∆dcl1 and ∆dcl2 mutant strains could not reveal the specific sizes of the vsiRNA products of each DCL. The abundance of TMV siRNA strongly decreased in both ∆dcl1 and ∆dcl2 mutant strains, but here vsiRNAs still retained lengths of predominantly 21 and 22 nucleotides, and the same 5 -terminal nucleotide preferences as in the wild-type strain ( Figure 4). Thus, in contrast with plants, in which each DCL generates siRNAs with a distinct size, single fungal DCL may produce heterogeneous sizes of siRNAs, although some size preferences may exist. Supporting this view, in Mucor circinelloides, DCL2 produces two different sizes (21 and 25 nucleotides) of antisense siRNAs that are associated with RNA silencing induced by the hairpin RNA-producing transgene [78]. The differential profiles of siRNAs between plant and fungal viruses may reflect substantial differences in replication processes between plant and fungal viruses; examples include differences in the replication site, type of membranes, and other cellular factors involved in viral replication. These differences may determine how DCLs gain access to and process virus-derived dsRNA substrates such as viral dsRNA replication intermediates and/or highly structured virus RNA genomes that are used for biogenesis of virus-derived siRNAs [79].
Although deletion of either the dcl1 or dcl2 gene markedly reduced TMV siRNA accumulation in F. graminearum, unexpectedly, deletion of both dcl1 and dcl2 genes restored TMV sRNA accumulation (Table 2 and Figure 1E). Moreover, TMV sRNAs accumulated in the double ∆dcl1-∆dcl2 mutant strain have similar size distributions, 5 -terminal nucleotide preferences, and distribution profiles along the viral genome to those accumulated in the wild-type strain (Figures 4 and 5). These results showed that F. graminearum DCL1 and DCL2 are involved in the production of TMV siRNAs. However, in the absence of both DCL1 and DCL2, there is an activation of a hitherto DCL-independent pathway to produce virus-derived sRNAs that have similar signatures to DCL-dependent vsiRNAs. As deletion of both dcl1 and dcl2 genes highly elevated TMV accumulation levels in F. graminearum [63], these highly accumulated DCL-independent TMV-derived sRNAs appear to be unable to mediate RNA silencing antiviral defenses, although they have similar features to the bona fide vsiRNAs produced by the DCLs. Similarly, a high accumulation of DCL-independent vsiRNAs with no antiviral function was also observed in dcl mutant strains of Cryphonectria parasitica and Colletotrichum higginsianum, but these DCL-independent vsiRNAs have different characteristics to DCL-dependent vsiRNAs [40,80]. Together, these observations indicate that in fungi, an alternative DCL-independent pathway processes virus-derived dsRNA precursors when functional DCLs are absent. It is still unclear what kind of fungal RNA endonuclease can alternatively cleave the viral RNAs. Previously, the presence of endogenous DCL-independent sRNAs had been identified in other fungi. In the basal fungus Mucor circinelloides, RdRP-dependent dicer-independent sRNAs are generated by RNase III-like protein specifically found in basal fungi and involved in the specific degradation of cellular mRNAs [81]. In N. crassa, DCL-independent siRNAs were found to mediate DNA methylation [34,82]. Further studies into the mechanisms underlying the biosynthesis of these DCL-independent vsiRNAs in fungi, and their biological significance, would be of interest.
Mycoviruses are mostly transmitted horizontally through hyphal anastomosis and vertically through sexual or asexual spores [35], but current knowledge has established that during fungal colonization, a plant and fungus can exchange macromolecules, including protein effectors, siRNAs, and nucleic acid parasites such as viruses and viroids [16,63,64,66,83]. From a broader evolutionary perspective, the finding that RNA silencing in fungi is functional and effective in conferring antiviral responses against plant viruses and viroids may indicate that fungal antiviral RNA silencing has also evolved to cope with cross-kingdom infection by the native molecular parasites of plants.