Characterizing the Roles of Cryphonectria parasitica RNA-Dependent RNA Polymerase-Like Genes in Antiviral Defense, Viral Recombination and Transposon Transcript Accumulation

An inducible RNA-silencing pathway, involving a single Dicer protein, DCL2, and a single Argonaute protein, AGL2, was recently shown to serve as an effective antiviral defense response in the chestnut blight fungus Cryphonectria parasitica. Eukaryotic RNA-dependent RNA polymerases (RdRPs) are frequently involved in transcriptional and posttranscriptional gene silencing and antiviral defense. We report here the identification and characterization of four RdRP genes (rdr1–4) in the C. parasitica genome. Sequence relationships with other eukaryotic RdRPs indicated that RDR1 and RDR2 were closely related to QDE-1, an RdRP involved in RNA silencing (“quelling”) in Neurospora crassa, whereas RDR3 was more closely related to the meiotic silencing gene SAD-1 in N. crassa. The RdRP domain of RDR4, related to N. crassa RRP-3 of unknown function, was truncated and showed evidence of alternative splicing. Similar to reports for dcl2 and agl2, the expression levels for rdr3 and rdr4 increased after hypovirus CHV-1/EP713 infection, while expression levels of rdr1 and rdr2 were unchanged. The virus-responsive induction patterns for rdr3 and rdr4 were altered in the Δdcl2 and Δagl2 strains, suggesting some level of interaction between rdr3 and rdr4 and the dcl2/agl2 silencing pathway. Single rdr gene knockouts Δrdr1–4, double knockouts Δrdr1/2, Δrdr2/3, Δrdr1/3, and a triple knockout, Δrdr1/2/3, were generated and evaluated for effects on fungal phenotype, the antiviral defense response, viral RNA recombination activity and transposon expression. None of the single or multiple rdr knockout strains displayed any phenotypic differences from the parental strains with or without viral infection or any significant changes in viral RNA accumulation or recombination activity or transposon RNA accumulation, indicating no detectable contribution by the C. parasitica rdr genes to these processes.


Introduction
Mycoviruses and other selfish genetic elements, e.g., transposons, are widely distributed throughout the filamentous fungi. A variety of mechanisms have been identified by which fungi restrict the accumulation and transmission of such elements [Reviewed in 1,2]. The genetic non-self recognition system, vegetative incompatibility, restricts transmission of mycoviruses [3][4][5][6] and potentially transposable elements [7,8] between genetically dissimilar individuals in a population. Silencing mechanisms act to control the transcription of transposable elements and accumulation of viruses. Demonstrated fungal transposon silencing mechanisms include repeat-induced point mutations (RIP) [9] and methylation [10]. RNA interference has been shown to restrict both transposon [11][12][13][14][15] and virus accumulation in filamentous fungi [16,17].
The RNA-silencing antiviral pathway is highly conserved among plants, invertebrates and fungi [18], consisting of two core components belonging to the Dicer-like and Argonaute nuclease families [19]. Dicer-like nucleases recognize and, with an associated RNase III-type activity, process viral double-stranded or highly structured RNA into small RNAs of 21-24 nts in length. With the aid of an Argonaute protein, these virus-derived small RNAs (vsRNAs) are subsequently incorporated into an effector complex called the RNA-induced silencing complex (RISC). One strand of the vsRNA is degraded and the remaining strand then serves to guide the effector complex to the cognate viral RNA that is then cleaved by an RNase H-like activity associated with the Argonaute protein. The role of RNA silencing as an antiviral defense mechanism in fungi was first demonstrated for the chestnut blight fungus Cryphonectria parasitica [16]. Although C. parasitica encodes two Dicer-like genes and four Argonaute-like genes, only Dicer 2 (dcl2) and Argonaute 2 (agl2) are required for antiviral defense [16,20]. DCL2 and AGL2 were also shown to contribute significantly to viral RNA recombination [20,21].
As clearly demonstrated for plants [14,22,23] and the nematode Caenorhabditis elegans [24][25][26], the Dicer/Argonaute-dependent antiviral RNA-silencing response is further amplified by host RNA-dependent RNA polymerases. Fungal-encoded RdRPs have been reported to play a role in transgene silencing [27][28], meiotic silencing [29][30], production of qiRNA induced by DNA damage [31], sex-induced silencing [32] and restriction of transposon accumulation [11][12][13][14]. The role of fungal RdRPs in the antiviral defense response is unknown. The primary goal of this study was to identify and characterize the RdRP genes in C. parasitica and examine their contribution to the RNA-silencing antiviral defense response and RNA silencing-mediated promotion of viral RNA recombination.

Fungal strains, growth condition and virulence assay
Wild-type C. parasitica strain EP155 (ATCC 38755) was used for generating rdr1 gene disruption strain Drdr1. Strain DK80, a mutant of EP155 disrupted in the ku80 gene for non-homologous end-joining DNA repair to promote integration of homologous DNA sequence [33], was used to generate all other RdRP genedisruption strains. All fungal cultures were maintained on PDA at 22uC. Viral infection of C. parasitica strains was achieved by transfecting viral RNA-coding strand transcripts into fungal spheroplasts by electroporation [34]. The in vitro growth of virus-infected and virus-free fungal mycelium of rdr gene knockouts and parental strains was assessed on PDA plates; colony diameters were measured after 7 days of growth. Asexual sporulation was measured in cultures grown on PDA for 14 days. Conidia were washed from plates with 1% Tween 20, subjected to serial dilution, counted in a hemocytometer and plated on PDA to test for viability.
Virulence assays were performed on dormant American chestnut tree stems as described by Hillman et al. [35], with six duplicate inoculations per fungal strain. Inoculated stems were kept at room temperature in a glass tank to maintain humidity. Cankers were measured after 21 days.
The sensitivity of wild-type and mutant strains to DNA mutagens was measured with a modified spot test [31]. Conidia were harvested with 0.8% NaCl, filtered through Miracloth, counted in a hemocytometer and diluted with 0.8% NaCl to a concentration of 10 8 /ml. Aliquots of the conidial stock solution were incubated on a shaker for 2 hrs at 60 rpm and 30 C with or without the DNA mutagens histidine (1.5 or 3.5 mg/ml) or hydroxyurea (2 or 4 mg/ml). The treated condial suspensions were taken through serial dilution and 10 ml, each containing 10 6 , 5610 3 , 5610 2 or 50 conidia, was spotted in duplicate on PDA plates containing the concentration of histidine or hydroxyurea used in the original treatment. Colony diameter was measured after 3 days of incubation at room temperature. Average values for the two replica plates were recorded.

Analysis of RdRP genome sequences, gene organization and phylogenetic relationships
The genome sequence assembly of the C. parasitica reference strain EP155, annotated with predicted or experimentally verified gene models and protein functions was retrieved from the U.S. Department of Energy's Joint Genome Institute (JGI; at http:// genomeportal.jgi-psf.org/Crypa2/Crypa2.info.html) and interrogated to identify putative RdRP genes. Additional sequence analyses of PCR-amplified cDNAs were performed by Macrogen (Rockville, MD).
Known or putative RdRP protein sequences from all major eukaryotic taxa were retrieved from GenBank or species genome websites. The sequences were entered into NCBI's Conserved Domain Search tool (at http://www.ncbi.nlm.nih.gov/Structure/ cdd/wrpsb.cgi), to identify the protein regions corresponding to the conserved RdRP domain, which were manually extracted and FASTA formatted. Alignments of the RdRP domain sequences were performed with the MUSCLE algorithm (Version 3.8 at http://www.ebi.ac.uk/Tools/msa/muscle/), using the default parameters. The protein alignments were imported into JalView [36] for inspection of alignments and construction of trees with the neighbor-joining method based on percent identity between sequences.

Generation of RdRP gene knockouts
Identified RdRP gene candidates were disrupted by replacing all or large parts of the rdr coding sequences (CDS) with selectable marker cassettes via homologous recombination, using the DNA transformation protocol developed by Churchill et al [37]. To increase the capacity for producing strains that contain multiple rdr gene disruptions, the Cre-loxP recombination system [38] was adapted to allow recycling of the limited number of selectable marker genes available for C. parasitica transformation [39]. The diagrams in Figs. S1-S3 show the details of the disruption schemes and origins of rdr1, rdr2 and rdr4 gene disruptions. The disruption scheme for rdr3 was previously described in Zhang et al. [39]. Overlap-extension PCR was used to generate the genedisruption fragments comprising the antibiotic resistance gene cassette and flanking sequences of the targeted rdr genes. Drdr1 was generated by replacing rdr1 with a cassette consisting of the trpC promoter from Aspergillus nidulans in front of the hygromycin B-phosphotransferase (hph) gene (Pro TrpC -hph) (Fig.  S1). Drdr2 was obtained by replacing a 1.5-kb rdr2 fragment with a 3-kb hph cassette (Fig. S2). Drdr3 was generated with a split marker approach using two fragments of a loxP-pro tubB -hph-loxP cassette as described in Zhang et al. [39]. Drdr4 was generated by replacing the rdr4 gene with the intact loxP-Pro tubB -hph-loxP cassette (Fig. S3).
To generate double rdr knockout strains, a neomycin (neo) gene cassette in pSK666 [20] was cloned into pKAES183 (provided by Dr. Christopher Schardl, University of Kentucky) by restrictionenzyme digestion and T4 enzyme ligation, to replace Pro tubBphleomycin with the neo gene behind the gpd promoter from A. nidulans, resulting in a loxP-Pro gpd -neo-loxP construct. The Drdr1/3 and Drdr2/3 double mutant strains were generated by disrupting the rdr1 or rdr2 genes with the loxP-Pro gpd -neo-loxP cassette in the Drdr3 strain background as described in Zhang et al. [39]. The Drdr1/2 double mutant strain was generated by disrupting the rdr1 gene with the loxP-Pro gpd -neo-loxP cassette in the Drdr2 strain background [described in Fig. 1B of 39]. The Cre-loxP system relies on the Cre recombinase to eliminate marker genes flanked by loxP sites, thus enabling multiple reuse of the same marker genes for generating serial knockouts [40]. As described in Zhang et al. [39] and included here for convenience, for use in recycling selectable marker genes in C. parasitica, a Pro gpd -Cre gene construct was cloned into pCPX-BSD1 [41], resulting in the construct pCPX-Cre, containing blasticidin as selectable marker. Using the method of Churchill et al. [37], the pCPX-Cre plasmid was transformed into Drdr1/3 and Drdr2/3 spheroplasts for Cre expression to eliminate selectable marker genes via recombination at the loxP sites flanking the hph or neo marker genes to produce marker gene-free strains Drdr1/3-M and Drdr2/3-M. Blasticidin was used at 600 mg/ml for transformant selection on regeneration media; although untransformed colonies also grew at this concentration, they did so at a slower rate than the pCPX-Cre-containing transformants. Therefore, putative transformants were selected by their larger colony size to screen for the loss of hygromycin and/or neomycin resistance by replicaplating on three different PDA media: medium containing 50 mg/ ml hygromycin PDA, medium containing 25 mg/ml neomycin PDA or antibiotic-free medium. The antibiotic-containing plates were screened for transformed colonies that displayed little or no growth, and the corresponding colonies on the antibiotic-free plates were subjected to single-spore isolation on PDA. Excision of the neo and hph genes was confirmed by PCR analysis using the same primer pairs used to confirm rdr gene knockouts.
Drdr1/2/3 triple knockout strains were engineered in two ways. One strain (F1) was constructed by generating a rdr1-disruption fragment containing Pro TrpC -hph by direct PCR amplification from the disrupted rdr1 gene in EP155-based disruption strain Drdr1, the product of which was then used to transform strain Drdr2/3-M. A second triple knockout strain, C1, was generated in the Drdr1/3-M strain background by disrupting the rdr2 gene with a Pro tubB -hph cassette without flanking loxP sites from pKAES173 (provided by Dr. Chris Schardl, University of Kentucky). Transformants were single spored and disruptions of all rdr genes were confirmed by PCR.
These extensive disruption studies took advantage of C. parasitic strain DK80 derived from reference strain EP155 by disruption of the ku80 gene for non-homologous end-joining DNA repair to promote homologous DNA recombination [33]. The rdr1 disruption mutant Drdr1 was generated in strain EP155 prior to the availability of strain DK80. No phenotypic differences have been observed between the DK80 strain and reference strain EP155, whether uninfected or following hypovirus infection. However, strain EP155 was used as the parental control strain for all experiments involving Drdr1 and DK80 was used as the parental control in experiments involving all other individual and multiple rdr gene disruption mutants. This resulted in the inclusion of both parental strains in many of the experiments described in this report.

Nucleic acid preparation and analysis
DNA extraction was performed with a small-scale method previously described by Spiering et al. [42]. RNA used for agarose gel analysis of viral full-length and defective interfering (DI) double-stranded RNAs was prepared as described by Sun et al. [20]. Total RNA for reverse-transcription real-time PCR (RT-qPCR) and RNA-seq analysis was extracted with Qiagen's (Valencia, CA) RNeasy kit. The Monster 1 st strand cDNA synthesis kit (Epicentre, Madison, WI) was used for cDNA synthesis from total RNA, following the manufacturer's protocol, and full-length cDNA copies of each rdr gene were amplified by PCR (Takara Tech Enzyme, Takara) and sequenced.
Transcript accumulation for selected genes was measured by one-step reverse-transcription real-time PCR (RT-qPCR), using AgPath-ID RT-PCR reagents (Ambion USA) and an ABI 7300 real-time PCR system. To guard against potential amplification of contaminating genomic DNA in the RNA samples, at least one primer or probe for each rdr gene was designed to span an intron/ exon boundary, identified by comparison of cDNA and genomic DNA sequences. Primer and probe design and general RT-qPCR methods were performed as described by Zhang et al. [43]. Calculations of transcript accumulation values were performed with the comparative threshold cycle method (DDCt) normalized against the b-tubulin gene transcript levels and calibrated against the normalized value in the wild-type strain. The primer/probe sets for the Dicer-2 and Argonaute-2 genes, dcl2 and agl2, respectively, were the same as described by Sun et al. [20]. Other primers used in this study are shown in Table 1.

Viral recombination assay
All rdr-knockout, DK80 and Ddcl2 strains were transfected with a replication-competent, GFP-containing viral vector Ctp40[2AEGFP] [44] by transfection with spheroplasts [34]. One week after the transfection, each infected strain was transferred to PDA and then sub-cultured to new PDA plates every 5 days for up to 6 transfers. Total RNA was isolated from sub-cultured colonies to monitor for presence of defective interfering viral dsRNAs and to serve as template to synthesize cDNA for PCR analysis with primer pair CHV1-2KF/CHV1-3KR [45] to monitor for maintenance of the EGFP coding domain in the viral vector RNA. GFP accumulation in infected fungal hyphae was also monitored under a fluorescence stereomicroscope (Model MVX10; Olympus) [21].

RNA-seq analysis of transposon expression
To identify and extract transposon sequences from the C. parasitica genome assembly at JGI (http://genomeportal.jgi-psf. org/Crypa2/Crypa2.home.html), the keywords of all gene models were queried with the following search terms: ''LINE,'' ''pogo'' and ''SINE.'' This search identified .500 gene models whose IDs and associated annotations were downloaded as csv files and sequences as FASTA files. Since these searches frequently resulted in repeated retrieval of the same sequences, the retrieved sequences IDs were manually checked and all duplicate sequences removed. Because many transposon copies in fungal genomes are highly similar in sequence-potentially affecting the read mappings in the RNA-seq analyses-all of the retrieved sequences were aligned with each other in BLASTN. Those sequences having .99% identity along .95% of their total length in the BLAST alignment were then combined as a single sequence consensus in JalView [36] for further analysis. In addition, sequences ,500 bp long were excluded from the analysis, resulting in a final set of 104 sequences. Of these, 71 were identified by their annotations in JGI or by BLASTX searches as class I transposable elements (i.e., retroelements or retrotransposons), 21 as class II (DNA) transposons and 12 sequences were similar to RNA-associated or hypothetical proteins (Table S1).
To quantify expression from each of the transposon sequences, total RNA was extracted from the rdr gene knockouts Drdr1/2/ 3(C1), Drdr1/2 and Drdr4 and the DK80 parental strain from mycelium grown on PDA for 7 days. Three biological replicates were used for Drdr1/2/3 and Drdr1/2 and analyzed separately by RNA-seq. One biological sample was used for Drdr4. The RNA was submitted to the Institute for Bioscience and Biotechnology Research DNA Sequencing Facility for construction of sequencing libraries containing cDNA fragments with an average length of 200 nucleotides. Paired-end reads sequencing data were acquired on an Illumina HiSeq 1000 instrument. On average, .40 million sequence reads were obtained per sample. The reads were uploaded into CLC Genomics Workbench 6.0 (CLCBio, Aarhus, Denmark) and mapped to the transposon sequences for RNA-seq analysis. The following (default) settings were used in the read mappings: a minimum length fraction of 0.9, minimum similarity fraction of 0.8 and maximum number of hits per read of 10; the minimum and maximum distances between paired reads were set at 100 and 250, respectively. Expression was estimated as reads per kb of coding sequence per million mapped reads (RPKM). The RNA-seq expression data from rdr gene knockouts and the DK80 control were used to calculate fold changes in expression by ratio-based analysis to identify differentially expressed transcripts from transposon sequences in each rdr gene knockout strain. Raw RNA-seq data have been deposited in the NCBI sequence read archive (SRA) under BioProject ID PRJNA258639.

Identification and features of RdRP-like genes in C. parasitica strain EP155
The C. parasitica genome sequence assembly generated by the U.S. Department of Energy's Joint Genome Institute for reference strain, EP155, was used as a resource for RdRP gene identification in this study. To identify RdRP gene candidates, the keywords in the annotated EP155 genome sequence were queried with ''RdRP'' as search term. This search yielded four RdRP-like sequences. In addition, a BLASTP search of the EP155 sequence was conducted, using QDE-1, the RdRP required for transgene silencing (Quelling) in Neurospora crassa [27,46], as query, which did not result in any additional sequences. The RdRP-like genes identified were designated rdr1 (scaffold_11:693898-697090 in JGI), rdr2 (scaffold_1:365214-369296), rdr3 (scaffold_5:3858759-3862568) and rdr4 (scaffold_4:4500867-4504921). The genomic sequences from each putative RdRP gene were then used to obtain cDNA amplicons by RT-PCR and sequencing. The PCRamplified cDNAs for rdr1, rdr2 and rdr3 gave a single product as judged by agarose gel analysis, and intron and exon positions in each of the three genes were identified by sequence comparison with the cognate genomic DNA sequences (Fig. 1). rdr1 had three introns and four exons, whereas rdr2 and rdr3 each contained only one intron and two exons. The location within the coding domain of the second intron in rdr1 corresponded to that of the only intron in rdr2 as judged by alignment of the encoded amino acid sequences relative to the exon-intron boundaries in the two genes. No such correspondence was seen for the only intron in rdr3.
Amplification of the full-length ORF in rdr4 gave multiple cDNA bands. To determine the reason for this potential diversity in rdr4 transcripts, the most prominent three bands were individually gel purified and sequenced. This analysis revealed that the variable banding pattern had most likely originated from alternative splicing events in the rdr4 gene. rdr4 contained three introns, and the first two introns were either spliced out or remained unspliced (Fig. 1). The four rdr genes have the following GenBank accession numbers: rdr1 (HF912382); rdr2 (HF912383); rdr3 (HF912384); rdr4 (HF912385).
Searches for conserved protein domains with the CD-Search Tool at NCBI (at http://www.ncbi.nlm.nih.gov/Structure/ bwrpsb/bwrpsb.cgi) revealed that RDR1, RDR2 and RDR3 each contained an RdRP-type conserved domain (pfam ID: pfam05183; E-values ,10 280 ). Interestingly, the rdr4 gene did not contain an intact RdRP domain; the predicted RDR4 protein sequence of the spliced transcripts contained in all cases a disrupted RdRP domain ( Fig. 2A) that lacked the DXDGD signature motif of eukaryotic RDRs (Fig. 2B). QDE-1 from N. crassa is involved in RNA silencing (Quelling) [27] and several key amino acid residues and domains required for its activity have recently been identified [47]. To see if the C.
parasitica RdRPs also contained these residues and domains, we conducted comparative sequence analysis of RDR1-RDR3 with QDE-1. Since rdr4 did not appear to encode a functional RDR protein, it was not included in the analysis. All three C. parasitica RDRs contained three invariant aspartic acid residues, corresponding to D1007, D1009 and D1011 in QDE-1 in its catalytic nih.gov/Structure/cdd/wrpsb.cgi) and are indicated by grey boxes; solid lines represent full-length proteins. E-values of each RdRP domain match and the range of amino acids included in the match are indicated next to and below the domain name, respectively. The longest predicted RDR4 protein is indicated here and contains an RdRP domain that is truncated due to a premature stop codon in its gene transcript and that lacks a conserved DPBB domain. B). The three C. parasitica RDRs containing complete RdRP domains were included in an alignment with QDE-1 from N. crassa, an RdRP involved in RNA silencing (Quelling); highly conserved DPBB domains and residues in QDE-1 [47] are indicated by boxes and are highlighted in red, respectively. doi:10.1371/journal.pone.0108653.g002 double-psi-b-barrel (DPBB2) domain [47] that constitute the DXDGD motif essential for catalyzing the formation of phosphodiester bonds during RNA synthesis [47]. Furthermore, all three RDRs contained three absolutely conserved residues present in the DPBB1 domain of QDE-1 (Fig. 2B). Including the DXDGD motif, ten highly conserved amino acid residues are present in the DPBB2 domain of QDE-1 [48], all of which were also present in RDR2; RDR1 shared nine and RDR3 eight out of the ten highly conserved residues in QDE-1. This analysis indicated that RDR1-RDR3 possess the active-site residues required for RDR activity; it also revealed that the RDR1 and RDR3 sequences were more divergent from QDE-1 than RDR2 in highly conserved RDR regions.
BLAST sequence alignments (all having E-values ,10 219 ) of the full-length RdRP sequences supported the above pattern of divergence among the three RdRPs: whereas QDE-1 aligned with RDR2 at 42% identity (59% similarity) along 1,125 amino acids, its alignment with RDR1 gave 40% identity (56% similarity) along 1,014 residues, indicating a somewhat greater level of divergence between RDR1 and QDE-1 (alignments not shown). RDR3 failed to give a contiguous alignment with QDE-1, and resulted in only 26% identity (43% similarity) along 340 residues of the longest aligned region. These analyses suggested that out of the three likely functional RdRP-like proteins in C. parasitica, RDR2 appeared to be the one most closely related to QDE-1. However, despite this relatively close relationship, RDR2 displayed only moderate (, 60%) similarity to QDE-1, suggesting the possibility of some functional divergence between the two proteins. None of the three C. parasitica RDRs contained an RNA-recognition motif (RRM; pfam ID: cd00590); such RRM domain was identifiable in N. crassa SAD-1 (meiotic silencing), but not in QDE-1, suggesting that the RRM motif is not essential for RNA silencing activity.

Sequence relationships of RdRP-like proteins from C. parasitica with other RdRP protein sequences
To gain some broader insight into the relationship of the C. parasitica RdRP-like sequences with putative and confirmed RdRP sequences from other species, a multiple sequence alignment of the RdRP-like C. parasitica sequences with those from a range of taxa from plants, protists, animals and other fungi was performed. The alignment was restricted to the predicted RdRP domains in each protein, and used to reconstruct a tree. This yielded several major clades, each represented by RDR sequences originating from a single kingdom (Fig. 3). Consistent with the general topology of RdRP trees in previous reports [49], RdRPs from animals grouped in a single well-defined clade, whereas those from fungi, plants and protists assembled in multiple clades, homogeneous for each organismal group. Interestingly, each of the three fungal RdRP clades contained an RdRP from N. crassa, i.e., QDE-1 (Quelling), SAD-1 (meiotic silencing) or RRP-3 (function unknown), a pattern that was closely tracked by the four RdRP-like sequences from C. parasitica. RDR2 was very closely affiliated with QDE-1, whereas the position of RDR1, although residing in the same (''QDE1'') clade suggested a more distant relationship with both RDR2 and QDE-1. RDR4 (included as two different predicted sequences inferred from alternative splicing and theoretical prediction) was affiliated with RDRs clustering with RRP-3, while RDR3 resided in another fungal clade that included SAD-1. These sequence relationships among the fungal RdRPs were reproducibly reconstructed with additional tree-building methods and evolutionary models (not shown). This analysis suggested distinct biological roles for the three functional C. parasitica RdRPs. For example, the affiliation with QDE-1 suggested involvement of RDR2 and possibly RDR1 in posttran-scriptional silencing, whereas clustering of RDR3 with SAD-1, an RDR involved in meiotic silencing [29], might suggest that RDR3 functions in a similar, but as-yet unreported, pathway in C. parasitica. A possible function for RDR4 was less clear, since it lacked a complete RdRP domain and no functional roles have yet been ascribed to its closest relative in N. crassa, RRP3. The analysis thus supported an early split of the four putative C. parasitica RdRP-like sequences, very likely preceding the speciation events that had resulted in the species shown in the tree.

Expression of rdr3 and rdr4 is induced upon infection with CHV1-EP713 and suppressed by the viral p29 protein
Expression of dcl2 and agl2 involved in the C. parasitica RNA silencing antiviral defense response is induced following hypovirus infection and suppressed by the virus-encoded RNA-silencing suppressor, papain-like protease p29 [20]. Thus, if one or more of the rdr genes were involved in anti-viral defenses, rdr expression would also be expected to be induced in response to viral infection and suppressed by p29. Therefore, the influence of infections by hypovirus CHV-1/EP713 and CHV-1/EP713 in which p29 was deleted (Dp29) on rdr gene expression was examined. Expression of rdr3 and rdr4 was significantly increased upon viral infection, and this response was greater when strain EP155 was infected by Dp29 (Fig. 4). By contrast, rdr1 and rdr2 expression showed only very minor responses to infection with CHV-1/EP713 or Dp29. Expression of rdr4 showed the strongest upregulation, similar to or even exceeding upregulation of dcl2 and agl2 in response to CHV1-EP713. These results indicated that expression of rdr3 and rdr4 was significantly induced by hypovirus infection, a response that was attenuated by the viral p29 RNA-silencing suppressor, thus exhibiting transcriptional responses that were somewhat similar to those of the anti-viral defense genes, dcl2 and agl2.

Virus-induced expression patterns for rdr3 and rdr4 are altered in Dagl2 and Ddcl2 strains
To investigate whether the expression of rdr genes may be affected by the deletion of the antiviral genes dcl2 or agl2, the strains Ddcl2 and Dagl2 generated in previous studies [20,21] were examined for rdr gene expression. Each strain was infected with CHV-1/EP713 or Dp29 and gene expression analyzed by RT-qPCR. Expression of rdr1 and rdr2 in virus-infected and virus-free strains Ddcl2 and Dagl2 (Fig. 5A-B) was largely similar to expression in strain EP155 (Fig. 4), suggesting little or no effect of the anti-viral gene deletions on expression of rdr1 and rdr2. In contrast, rdr3 expression was increased $10-fold upon virus infection in both Ddcl2 and Dagl2, much greater than the 2 to 6fold increase in expression of rdr3 in virus-infected EP155. rdr4 expression was only marginally (#2-fold) increased in both Ddcl2 and Dagl2 in response to CHV1-EP713 infection, down from ,20-fold upregulation in CHV1-EP713-infected EP155. The level of upregulation of rdr4 expression in Dp29-infected Dagl2 was largely similar to that in Dp29-infected EP155, whereas deletion of dcl2 markedly decreased rdr4 upregulation to ,3fold, compared to the almost 30-fold upregulation of rdr4 in Dp29-infected EP155. Expression of rdr1 and rdr2 did not exhibit any detectable changes. These results suggested interactions between rdr3 and rdr4 expression and the induction of the dcl2/agl2 RNA silencing pathway in response to virus infection.

Disruption of RdRP genes and testing of contributions to the RNA silencing antiviral defense response
To further investigate the potential roles of the C. parasitica RdRP genes in the RNA silencing antiviral defense response, all four RdRP genes were disrupted individually and in combination. C. parasitica strains carrying disruptions of the antiviral genes dcl2 and agl2 became severely debilitated following hypovirus infection [16,20]. So it was a reasonable expectation that disruption of an RdRP gene involved in the RNA silencing antiviral defense pathway would likewise result in similar readily detectable phenotypic changes. The single Drdr disruption strains each showed growth phenotypes that were essentially identical to the phenotype of the parental strains both in the absence and presence of infecting hypovirus CHV-1/EP713 (Fig. 6).
In view of these results and to account for the possibility that the RdRP genes have some functional redundancy, a series of doubleand triple-disruption mutant strains was generated as described in the methods. rdr4 was not included in this analysis, since, as shown above, its predicted protein product lacks a functional RdRP domain. Double and triple Drdr1/2/3 knockout strains also exhibited normal growth and colony morphology (examples shown in Fig. 6). Additional analysis revealed no reduction in growth of the multiple rdr knockout strains in a virulence assay on inoculated dormant chestnut stems relative to the parental strain or any change in the production of asexual spores (data not shown). N. crassa qde-1 mutant strains showed sensitivity to DNA damaging agents [31]. In contrast, the C. parasitica double and triple rdr mutant strains were as sensitive as the wild-type strain to increasing concentrations of histidine and hydroxyurea ( Table 2). The combined results for all single, double and triple rdr knockout strains revealed no essential role for the rdr genes in fungal growth, asexual sporulation, virulence or response to DNA damage-related stress.
CHV-1/EP713 infection of the double and triple rdr mutant strains resulted in the standard phenotypic changes that were indistinguishable from those of the CHV-1/EP713-infected parental strains (examples shown in Fig. 6). In contrast, growth of the CHV1-EP713-infected Ddcl2 strain used as a control for deletion of antiviral genes in this experiment exhibited a highly debilitated growth phenotype (Fig. 6), as previously described [16]. Moreover, viral RNA accumulation levels were not significantly different between parental and rdr mutant strains (data not shown). Based on the combined results, the C. parasitica rdr genes appear to play no significant role in the RNA silencing antiviral defense response.

Contribution of host RdRPs to viral recombination
The C. parasitica RNA-silencing antiviral defense pathway was shown to contribute to viral RNA recombination [reviewed in 50]. Defective interfering (DI) RNAs that are readily formed in CHV- 1/EP713-infected strain EP155, fail to form in either Ddcl2 or Dagl2 mutant strains [20,21]. Moreover, non-viral sequences that are rapidly deleted from recombinant hypovirus RNA virus vectors in wild type C. parasitica are stably maintained over prolonged periods of sub-culturing in the Ddcl2 and Dagl2 strains [20,21]. The collection of rdr gene-disruption mutant strains was tested for their ability to support viral RNA recombination. Mutant and control strains were infected with hypovirus CHV-1/ EP713 and independently with recombinant hypovirus Ctp40[2AEGFP] that expresses the enhanced green fluorescent protein (EGFP). The infected strains were sub-cultured every 5 days and monitored for fluorescence, maintainence of the EGFP sequence and production of DI RNAs. Fluorescence disappeared in all independent rdr gene-disruption mutants and in the Drdr1/ 2/3 triple mutant within three subcultures, similar to what was observed for the control-infected parental strain. In contrast, fluorescence was retained in the infected Ddcl2 strain for extended periods of sub-culturing. PCR analysis of the extracted viral RNA confirmed the loss of the recombinant viral EGFP gene sequence in the Drdr mutant strains as early as the first sub-culture and persistence of the EGFP gene sequence in the Ddcl2 strain. Similary, DI RNAs were readily detected in the virus-infected parental strains and in the individual and triple rdr gene disruption strains, but not in the virus-infected Ddcl2 strain (data not shown).
Thus, these extensive sets of infection sub-culturing experiments (performed in triplicate) failed to reveal any observable contribution of the rdr genes to RNA silencing-mediated promotion of viral RNA recombination.

Contribution of RdRPs to transposon silencing
We next investigated whether knocking out one or more of the rdr genes affected the expression of transposable elements. Because the C. parasitica genome contains .100 transposon and transposon-like consensus sequences of .500 bp (see Methods), RNA-seq was performed to investigate whether global transposon expression was altered in the rdr gene knockouts. To identify and retrieve transposon sequences, we used a series of queries to search the annotated C. parasitica genome on the JGI website (see Methods), which identified 104 known or putative transposable elements. Illumina sequencing reads representing the entire transcriptome in Drdr1/2/3, Drdr1/2 and Drdr4 and the parental DK80 strain were then used in RNA-seq mappings to these transposon sequences and expression analysis in CLCBio's Genomics Workbench. About 0.1% of the total reads in each strain mapped to the transposon sequences (not shown), and RNAseq detected expression for 60 of the transposon sequences in at least one of the strains. The RNA-seq results indicated that expression of these transposon sequences was very variable: some  [20,21] and were also measured in parallel here. The identities of the transcript being measured are indicated at the bottom of the histogram. The fold differences in transcript accumulation are indicated on the Y axis and were calculated using the comparative threshold cycle (DDCT) normalized against tubb transcript levels and calibrated against the normalized value in the wild-type strain, with standard deviations, based on three independent measurements of two independent RNA preparations, indicated by the error bars. Transcript levels for uninfected C. parasitica were set to a value of one. doi:10.1371/journal.pone.0108653.g004 sequences yielded ,10 mapped reads, whereas others resulted in .9,000 mapped reads. However, only ,20 transposable elements displayed significant expression (i.e., .100 mapped reads per sequence), none of which showed significant (i.e., $2 fold) up-or down-regulation in any of the rdr gene knockout strains (results for Drdr1/2/3 and Drdr4 are shown in Table S1). Thus, disruption of the rdr genes had no apparent effect on transposon transcript accumulation.

Discussion
Molecular genetic characterization of the RNA silencing components encoded by C. parasitica has clearly demonstrated that the role of RNA silencing as an antiviral defense response observed in plants and invertebrates extends to the fungi. However, only a subset of the RNA silencing apparatus, a single Dicer protein, DCL2, and a single Argonaute protein, AGL2, are required for this important function. The antiviral defense response functions independently of the additional Dicer gene, dcl1, and three additional Argonaute genes, agl1, agl3 and agl4, encoded by this fungus. The extensive analysis of the C. parasitica RdRP genes described here failed to find any evidence for contributions by any of these components to the antiviral defense response.
The four C. parasitica rdr genes fit into the three clades of RdRPs found in most filamentous fungi [51] that correspond to the three N. crassa RdRP genes. The C. parasitica rdr4 gene clustered in the clade containing N. crassa RRP-3, for which no function has been identified. Interestingly, rdr4 displayed alternative splicing, resulting in each case in transcripts whose products lacked an intact conserved RdRP domain, including the essential DXDGD motif. Thus, rdr4 appears to be a non-functional RdRP pseudogene. In this regard, there is considerable evidence for the loss and expansion of RNA silencing genes in fungi [52], e.g., the apparent degeneration of the Aspergillus nidulans QDE-1  ortholog [51] and truncation of A. nidulans Dicer DCL-1 and Argonaute SMS-2 orthologs [17].
Of the three C. parasitica RdRP genes with intact RdRP motifs, rdr3 clustered with N. crassa SAD-1, required for meiotic silencing of unpaired DNA (MSUD), while both rdr1 and rdr2 clustered with N. crassa qde-1, involved in vegetative silencing or Quelling. Since the single Dicer gene dcl2 and single Argonaute gene agl2 required for the antiviral defense response in C. parasitica are orthologs of the N. crassa Dicer (dcl-2) and Argonaute (qde-2) genes involved in Quelling, the rdr1 and rdr2 genes would logically be the most likely candidates to be involved in possibly amplifying the antiviral response.
Examination of RdRP gene expression levels in response to hypovirus infection showed that rdr1 and rdr2 transcript levels did not increase following virus infection as was observed for the dcl2 and agl2 antiviral defense genes. Surprisingly, the accumulation of rdr3 and rdr4 transcripts did significantly increase in response to virus infection (Fig. 4). Moreover, as reported for dcl2 and agl2 [20,21], rdr3 and rdr4 expression was further increased following infection of the Dp29 virus lacking the suppressor of RNA silencing p29 (Fig. 4). Altered patterns of rdr3 and rdr4 expression were also observed in the Ddcl2 and Dagl2 mutant strains (Fig. 5). Interestingly, expression of the N. crassa rdr4 ortholog rrp3 increased 30-fold in response to dsRNA, leading Choudhary et al. [53] to propose that rrp3 may play a role in the formation of viral or transposon dsRNA.
While the transcript accumulation results indicated some level of interaction at the expression level between the antiviral defense genes and at least rdr3 and rdr4, supporting evidence for a role of the rdr genes in the antiviral defense response was not forthcoming from extensive disruption analysis of the rdr genes. No alterations in either virus-induced symptoms or in virus accumulation levels, as observed for dcl2 and agl2 disruption mutants, were detected in single or multiple rdr gene disruption mutant strain including the triple rdr1, rdr2, rdr3 mutant. Thus, we conclude that C. parasitica is able to mount a robust RNA-silencing antiviral defense response without contributions from the encoded rdr genes. This clearly contrasts with the observations that RdRPs serve to amplify the RNA silencing antiviral defense response in plants [14,22,23] and in C. elegans [24][25][26]. However, the very large induction in C. parasitica dcl2 expression in response to CHV1-EP713 infection [20,21] may preclude the need for RdRP amplification of the antiviral defense signal.
We recently reported that the C. parasitica Dicer and Argonaute proteins involved in antiviral defense also contribute to viral RNA recombination to produce viral defective interfering (DI) RNAs and to delete non-viral RNA sequences from viralbased expression vectors [20,21]. DI RNA formation and vector RNA stability was indistinguishable in parental, individual rdr mutant and triple rdr mutant strains. Thus, the rdrs also appear not to contribute to this RNA-silencing antiviral defense-related effect.
The accumulation of transposable element transcripts and fragments have been reported to be modulated by RNA-silencing pathways in several filamentous fungi [12,13,15,32,54,55], and specific roles for RdRPs in this process have been reported for Cryptococcus neoformans [12,32], Mucor circinelloides [13] and N. crassa [54]. Illumina sequencing of C. parasitica small RNAs isolated from wild type and Ddcl2/Dagl2 strains identified the presence of DCL2/AGL2-derived transposable element-specific siRNAs (data not shown), indicating targeting of TE RNA by the antiviral defense RNA-silencing pathway. However, no significant increase in TE transcript accumulation was observed in RNA-seq analysis of the rdr knockout strains, including the triple knockout Table 2. Colony diameter (cm) in DNA-damage sensitivity assay measured after 3 days of growth.  (Table S1). In this regard, Nolan et al. [11] reported that silencing of the introduced LINE1-like transposon Tad in N. crassa required one of the two Dicer genes and Argonaute gene qde-2, but surprisingly, not the RdRP gene qde1. The mechanism leading to the generation of TE-specific siRNAs in C. parasitica requires additional investigation.
The determination that C. parasitica rdr genes do not play a significant role in the RNA-silencing antiviral defense response, the primary aim of the study, is important information with bearing on future investigations. Disruption of individual and combinations of rdr genes also resulted in no observable changes in growth, colony morphology, asexual sporulation or virulence on chestnut. Similar results were observed for individual argonaute disruption mutants and individual and double dicer genedisruption mutants in the absence of virus infection. Thus, the RNA-silencing components in C. parasitica also appear not to contribute to developmental, metabolic or morphological functions during vegetative growth, as has been reported for dcl2 and rdr3 disruption mutants of Trichoderma atroviride [56]. Reports of a role for RNA-silencing pathways in stress responses have been mixed. The N. crassa gdq-1 rdr and dicer gene mutant strains were reported to be more sensitive to histidine and DNAdamaging agents, such as hydroxyurea [31], while C. neoformans RNA-silencing mutants showed no alterations in stress responses [12]. The C. parasitica individual and combined rdr mutants and Ddcl2 failed to show any altered response to histidine or hydroxyurea ( Table 2) and appear not to be involved in stress responses. However, the extensive characterization reported here does not eliminate possible roles for the C. parasitica rdrs in cosuppression or meiotic silencing, a determination that will require the development of additional experimental capabilities with this fungus. Figure S1 Disruption of C. parasitica rdr1 in EP155. The 4957 bp fragment (850 bp left arm + rdr1 gene 3763 bp + 344 bp right arm) is replaced by pro-trpC-hph cassette (2172 bp). The primer pair R1f (310 bp upstream of the disruption fragment) and R1r (727 bp downstream of the disruption fragment) was designed to confirm the disruption yielding a 6.0 kb fragment from the wildtype strain and a 3.2 kb fragment from Drdr1. (TIF) Figure S2 Disruption of C. parasitica rdr2. Drdr2 was generated by the replacement of the 1.5 kb fragment in the rdr2 gene with a 3-kb hph cassette. The disruption confirmation primer R2f/R2r pair amplified the intact rdr2 gene (,4.1 kb) in the wildtype strain and a 5.7 kb fragment in Drdr2.