The C. parasitica strain EP155/2 (ATCC 38755), used as the wild-type strain, and its isogenic hypovirus-CHV1-713-containing strain, UEP1, were cultured on PDAmb plates under constant low light at 25°C (Kim et al., 1995). The culture conditions and methods for preparation of the primary inoculum for liquid culture were described previously (Kim et al., 1995). The collected mycelia were lyophilized and stored at -70°C until use (Powell and Van Alfen, 1987). Strains used for the analysis of DNA methylation were the CpBck1-null mutant (TdBCK1), the sectored progeny of the CpBck1-null mutant (TdBCK1-S1), and the complemented strain of the sectored CpBck1-null mutant strains (TcBCK1-S1) from our previous study (Kim et al., 2016). The wild-type EP155/2 strain was included for comparison purposes.

Genomic DNA was extracted using lyophilized mycelia (Kim et al., 1995) and was further purified using a G25 column (GE Healthcare, Little Chalfont, BUX, United Kingdom; Gobbi et al., 1990;Asanuma et al., 2013). Quantification of 5-methylcytosine (5mC) was performed using a MethylFlash^TMMethylated DNA Quantification Kit including ME4 as a control (50% of 5-methylcytosine; Epigentek, New York, NY, United States), as described previously (Simmons et al., 2013).

Whole-genome bisulfite sequencing was performed as described previously (Kulis et al., 2012). Briefly, genomic DNA (5 μg) was mechanically sheared into fragments, and the fragments were end-repaired for ligation to a methylated adapter with a single-base ‘T’ overhang using Truseq DNA library preparation kit (Illumina, San Diego, CA, United States). The ligated products were size-selected by agarose gel electrophoresis, and bisulfite conversion of the fragments was performed as described previously using an EpiTect Bisulfite Kit (Qiagen, Manchester, MCH, United Kingdom). The treated product was amplified by PCR using a uracil-tolerant proofreading enzyme (Pfu Turbo Cx, Stratagene, La Jolla, CA, United States). Libraries were paired-end sequenced using the Illumina HiSeq2000 platform. Raw sequencing data were checked for quality using the FastQC program, and alignment and mapping with the reference genome data of C. parasitica¹ were performed to identify the methylation of cytosine residues using Bismark and Bowtie2 software, as described previously (Xiang et al., 2010). The degree of DNA methylation was indicated by the methylation level and density, which were defined as the number of mC reads divided by the number of total reads covering the site and as the number of mC sites per 10 kb, respectively.

Genomic DNA of C. parasitica was fragmented using restriction enzymes that cut outside target regions of interest and was subjected to manual bisulfite conversion (Colella et al., 2003; Clark et al., 2006). Based on the WGBS data, the target regions, which read count was more than 100 and methylation occurred at all context in all four samples, were randomly selected and analyzed. Amplified products were cloned using the pGEM-T Easy vector (Promega, Madison, WI, United States), and at least 18 independent clones were sequenced to analyze the target region. Primer pairs for bisulfite-PCR validation are shown in Supplementary Table S1.

The genome database of C. parasitica was screened for C5-DNA-methyltransferase genes. Two putative C5-DNA-methyltransferase genes CpDmt1 and CpDmt2 were selected for further analysis. PCR amplification of CpDmt1 was performed using the primers CPDMTC5_2549_F1 and CPDMTC5_2549_R2 and that of CpDmt2 was performed using the primers CPDMTC5_1891_F1 and CPDMTC5_1891_R2 (Supplementary Table S1). The resulting 5.6-kb PCR amplicon of CpDmt1 and 7.3-kb amplicon of CpDmt2 were cloned separately into the pGEM-T Easy vector and sequenced using the dideoxynucleotide method with universal and synthetic oligonucleotide primers.

To obtain the cDNA clones of CpDmt1 and CpDmt2, PCR was performed using reverse transcriptase (RT-PCR) with primers CpDmt1-mF1/CpDmt-mR1 and CpDmt2-mF1/CpDmt2-mR1 at nucleotide positions (nt) -8 to 14/3,272 to 3,291 and (nt) -8 to 12/3,796 to 3,815 (relative to the start codon), respectively (Supplementary Table S1). The resulting 3.3- kb and 3.8-kb cDNA amplicons were cloned and sequenced.

The phylogenetic relationships of CpDmt1 and CpDmt2 with representative fungal cytosine-specific methyltransferase domains were analyzed using MEGA7 (Kumar et al., 2015). The maximum likelihood method based on the JTT matrix-based model was used to estimate the evolutionary history (Jones et al., 1992).

The protein domain features were predicted and compared among the selected fungal methyltransferases using batch CD-search of the NCBI Conserved Domain Database CDD (Marchler-Bauer and Bryant, 2004; Marchler-Bauer et al., 2015).

To examine the expression levels of target genes, quantitative RT-PCR was performed (Park et al., 2012). Briefly, a GeneAmp 7500 sequence detection system (Applied Biosystems, Foster City, CA, United States) using a SYBR green mixture RT kit (Applied Biosystems) was applied. 1 μg of total RNA was treated with RNase-free RQ1 DNaseI and then was reverse transcribed. Real-time PCR was performed on 2 μL of reverse-transcribed cDNA with 1 μL of each forward and reverse primer at 150 nM concentration. The glyceraldehyde-3-phosphate dehydrogenase gene (Gpd) was used as an internal control. Analyses were conducted at least two independent RNA preparations, in triplicate for each transcript, with primers specific for Gpd and the target genes. Primer pairs for each gene are shown in Supplementary Table S1. Transcript abundance, relative to the amount of Gpd, in the sample was calculated based on the fold change expression of target genes normalized to the internal control Gpd (Parsley et al., 2002). RNA was extracted from liquid cultures as described previously (Kim et al., 1995).

The replacement vectors pDDMT1 and pDDMT2, which were designed to favor double-crossover integration events, were constructed. In the replacement vector pDDMT1, the hygromycin phosphotransferase gene cassette (hph) was inserted between sites nt 15 and 1,968 of the CpDmt1 gene relative to the start codon and flanked by approximately 1.7 and 2.0 kb of 5′ and 3′ sequences, respectively. The 5.6-kb CpDmt1 disruption cassette was then used to transform the virus-free EP155/2 strain. In a similar way, in the replacement vector pDDMT2, the hygromycin phosphotransferase gene cassette (hph) was inserted between sites nt 1 and 3,367 of the CpDmt2 gene relative to the start codon and flanked by approximately 2.0 and 1.9 kb of 5′ and 3′ sequences, respectively. The 7.3-kb CpDmt2 disruption cassette was then used to transform the virus-free EP155/2 strain.

Functional complementation of the CpDmt1-null mutant using a wild-type allele was performed. The complementing vector pCDMT1 was constructed by insertion of SpeI/SacII-digested pCDMT1 carrying a 5.6-kb fragment containing the full-length CpDmt1 gene into SpeI/SacII digested pBSSKG plasmid containing the geneticin resistance cassette of pSilent-Dual1G (pSD1G) in pBluescriptII SK(+) (Nguyen et al., 2008). The resulting vector was then used to transform the CpDmt1-null mutant. Functional complementation of the CpDmt2-null mutant was carried out in a similar fashion. The complementing vector, pCDMT2, was constructed by inserting HindIII/SpeI-digested pCDMT2 carrying a 7.0-kb fragment containing the full-length CpDmt2 gene into the pBSSKG plasmid digested with HindIII and SpeI, followed by transformation into the CpDmt2-null mutant.

Genomic DNA was isolated from C. parasitica as described previously (Churchill et al., 1990). DNA (10 μg) was digested with the appropriate restriction enzyme, blotted onto a nylon membrane, and hybridized with radioactive-labeled probes (Sambrook et al., 1989).

The phenotypic and molecular characteristics of the CpDmt1- and CpDmt2-null mutants were examined with the comparison of wild-type EP155/2 and the hypovirulent UEP1 strains. Phenotypic changes, such as growth rate, pigmentation, conidiation, and mating capability, were measured as described previously (Powell and Van Alfen, 1987; Kim et al., 2002). Morphological characteristics were examined on PDAmb medium.

All statistical analyses were performed in SPSS version 22.0 (SPSS Inc., Chicago, IL, United States). Pearson’s Chi-square test was used to evaluate the difference in methylation frequency between strains. One-way ANOVA was used to analyze the significant differences of relative gene expression.

Quantification of global DNA methylation was performed using a MethylFlash^TM Methylation DNA Quantification Kit (Epigentek). Total DNA from the wild-type EP155/2, CpBck1-null mutant (TdBCK1), sectored progeny of the CpBck1-null mutant (TdBCK1-S1), and the complemented strain of the sectored CpBck1-null mutant strains (TcBCK1-S1) were prepared (Kim et al., 2016). As shown in Supplementary Figure S1, DNA methylation was detected in all four strains. Interestingly, there were differences in global DNA methylation among strains; the optical density of methylated DNA ranged from 0.150 to 0.231 among samples, in the order of wild-type EP155/2, TdBCK1, TcBCK1-S1, and TdBCK1-S1. Interestingly, the methylation levels of the mutant strains TdBCK1, TcBCK1-S1, and TdBCK1-S1 were significantly lower than that of the wild-type (p < 0.05, one-way ANOVA).

Whole-genome bisulfite sequencing enables analysis of the methylation state of all cytosine residues in an individual DNA sequence (Clark et al., 2006; Xiang et al., 2010; Jeon et al., 2015). Thus, WGBS of genomic DNAs extracted from mycelia of the wild-type, TdBCK1, TdBCK1-S1, and TcBCK1-S1 strains was performed by next-generation sequencing (NGS) using an Illumina HiSeq 2000 system. Read pairs were subsequently mapped to the corresponding reference genome using BisMark (Nuss et al., 2010; Krueger and Andrews, 2011). These procedures resulted in 49–57 million reads corresponding to 9.8, 10.6, 11.5, and 10.6 Gb high-quality sequences, respectively, with average map rates of 67–81% in read alignment (Supplementary Table S2). This represented approximately 75× to 99× coverage of the C. parasitica genome, the reference sequence of which has an estimated size of 43.9 Mb. Lambda DNA, which lacks DNA methylation, was included in the bisulfite treatment to estimate the conversion rate (Remnant et al., 2016). In our experiment, the conversion rate was at least 99.4%. Additionally, to minimize the bias that may result from the overrepresentation per site, sites read more than 500 times or fewer than five times were excluded from the analysis, which allowed the methylation levels of individual sites to be determined with reasonable confidence. The average map rates of cytosine sites analyzed after the coverage cutoff (5–500) were 71.9, 66.7, 80.1, and 75.5% in read alignment for the wild-type, TdBCK1, TdBCK1-S1, and TcBCK1-S1 strains, respectively.

Whole-genome bisulfite sequencing data were validated by cloning and sequencing PCR products from bisulfite-treated DNA and analysis by CyMATE for two representative [scaffold_13: 32825-32935 and scaffold_13: 33015-33109] loci predicted to carry methylation (Supplementary Table S3). Comparison of two independent datasets showed that over 90% of the methylated cytosine (mC) sites identified in our WGBS overlapped with mCs on CyMATE analysis, suggesting the sensitivity and reliability of the WGBS performed here.

The global methylation level was estimated by dividing the amount of methylated cytosine by the total amount of cytosine detected in each sample (mC/total C). The global methylation level of the wild-type was 3.9%, which was low but sufficiently significant (p < 0.001, Chi-square test) and within the range in other fungal species (2–6%). However, changes in the global methylation levels were observed in mutant strains. The global methylation level of TdBCK1-S1 was lowest, and only 1.0% of detected cytosines were methylated. Interestingly, the complemented strain TcBCK1-S1 showed restoration to a level of 2.1%. The parental mutant strain TdBCK1 showed an intermediate level of 2.6%. However, global DNA methylation levels are only an indirect indicator of the mCs in the genome. Therefore, further analyses were conducted to determine the variation in number of mC sites among the four strains (Supplementary Table S4).

In total, 1,410,293 cytosine sites were identified as mC sites in the wild-type EP155/2, accounting for approximately 6.4% of all genomic cytosine sites that experienced methylation. However, more mC sites were observed in the mutant strains. The parental TdBCK1 strain showed the most mC sites, with a total of 1,694,399 sites, accounting for approximately 7.7% of all genomic cytosines. Considering the least average map rate of 66.7% (Supplementary Table S2) and the most mC sites among the four strains, the increased number of mC sites in the TdBCK1 strain was dramatic. The complemented TcBCK1-S1 and sectored TdBCK1-S1 strains showed an intermediate number of mC sites (1,576,353 and 1,575,942, respectively). Considering the average map rates of 75.5 and 80.1% covered in the TcBCK1-S1 and TdBCK1-S1 strains, respectively, which were higher than 71.9% in the wild-type strain, the increased number of mC sites may not be as dramatic as that of TdBCK1. However, these two mutant strains still showed an increased number and proportion of mC sites compared to wild-type. Thus, it is interesting that the wild-type EP155/2, which contains the highest global methylation level among the four strains, showed the fewest number of mC sites. In addition, the sectored progeny (TdBCK-S1) exhibited marked genome-wide DNA hypomethylation but an increase in methylation sites. C. parasitica showed cytosine methylation in any DNA context, including symmetrical CG and CHG (where H = A, T, or C) and asymmetrical CHH, as in other species of ascomycetes (Zemach et al., 2010; Aramayo and Selker, 2013). The methylation proportions representing the ratio among all three classes of potential methylation site in the wild-type were 71.6, 21.8, and 6.6% for CHH, CHG, and CG, respectively (Supplementary Table S4). These proportions were quite different from those expected based on random frequencies of 56, 19, and 25%, respectively, and indicated strong methylation preference for CHH and CHG sites. However, methylation proportions of all three classes of potential methylation site were similar in all four strains. They were, in order, CHH (68.7–71.6%), CHG (21.8–22.6%), and CG (6.6–8.9%); therefore, they represented epigenetic homogeneity with respect to all three classes of potential methylation site. The global methylation level of CHH and CHG sites in the wild-type strain was 4.3%, while those of CHG and CHH sites were 2.7 and 3.0% in the parental TdBCK1, 1.1 and 1.1% in TdBCK1-S1, and 2.3 and 2.4% in TcBCK1-S1, respectively. The global methylation levels of CG sites were 0.09, 0.07, 0.05, and 0.06% for the wild-type EP155/2, TdBCK1, TdBCK1-S1, and TcBCK1-S1 strains, respectively, which were also markedly lower than their corresponding global methylation levels.

As shown in Supplementary Figure S2, mC sites in the genome of the wild-type were not evenly distributed, but rather formed relatively densely methylated domains. Moreover, when compared with the gene maps showing the density of putative genes, these densely methylated regions were located mostly around intergenic regions and gene-poor regions across contigs (Figure 1). Interestingly, the clustering pattern of mC sites was similar among all four strains except for the domain in supercontig 4, which showed the appearance of a new densely methylated domain in the mutant strains compared to the wild-type strain (Supplementary Figure S2). Considering the differences in global methylation level and the number of mC sites in the wild-type strain, the differences in DNA methylation patterns among the strains were ascribed to changes in both the methylation level (defined as the number of mC reads divided by the number of total reads covering the site) of each individual mC site and the density (defined as the number of mC sites per 10 kb) of mC sites. However, similarities in the patterns of mC site clusters among the strains suggest that changes in mC sites are restricted to areas around the mC site clusters. The differentially methylated domain in supercontig 4 of the mutant strains is interesting because it is not located in the gene poor region and was identified as a new cluster. Thus, we examined the differentially methylated genes in this specific domain (supercontig 4: 1,530,000–1,540,000) and found that four predicted genes (Supplementary Table S5). Among these four predicted genes, one was suggested to be a transmembrane transporter and three are without any known protein domain. Although these results prompt further investigation, we hypothesize there may be significant changes in the metabolites of the mutant strains from those of the wild-type strain. Interestingly, adjacent to the mutant-specific differentially methylated region, there are highly methylated regions specific for the TdBCK1 strain. Sequence analysis of the three most highly methylated regions specific for TdBCK1 (supercontig 4: 1,840,000–1,850,000, 1,870,000–1,880,000, and 1,920,000–1,930,000) revealed six predicted genes. Among these, one was identified as a gene containing a domain involved in carbohydrate metabolic process while the others did not have any specific domain (data not shown).

The distribution of methylation levels for each mC site was analyzed. As shown in Figure 2, the methylation levels of individual mC sites remained less than 70% across samples, with a strong preference for ≤40%. Although the proportions of mC sites with different methylation levels were similar across samples, i.e., a sharp decrease from the ≤10% level to the next level of 20% and then a gradual decrease to 70%, the wild-type strain showed a larger proportion of the ≥20% methylation level (Figure 2A) compared to the other three strains (Figures 2B–D). These results were in agreement with the results of the global methylation levels. Moreover, the methylation levels of CG sites were mostly ≤10% among all strains. Similar results were found in Uncinocarpus reesii and N. crassa (Zemach et al., 2010; Aramayo and Selker, 2013). Thus, methylation occurred less frequently in the CG sites and the methylation level at individual CG sites, if any, showed a strong preference for ≤10% (Supplementary Table S4), which resulted in markedly lower global level of methylation at the CG sites than at the CHH and CHG sites. Additionally, considering the fewer number of mC sites, low methylation levels, and relatively small changes in the methylation levels in the CG sites among strains, hypomethylation of mC sites in mutant strains occurred mainly in the CHH and CHG rather than the CG context.

The distribution of mC sites to different genomic features such as intergenic regions, genes (exons and introns), and upstream (1.5 kb, relative to the putative start codon) and downstream (1.0 kb, relative to the putative stop codon) of genes were analyzed by calculating and comparing DNA methylation densities of the whole genome (Supplementary Table S6). Compared to the intergenic region encompassing 70% of methylation sites, the proportion of mC sites in genic regions in the wild-type was only 30%, which was significantly low. Interestingly, TdBCK1 and its sectored progeny, TdBCK1-S1, showed the decreased proportions of mC in intergenic regions of 62 and 61%, respectively. The complemented strain, TcBCK1-S1, showed an intermediate proportion (65%) of mC sites in intergenic regions. Within genic regions of the wild-type strain, methylation was found in both coding and non-coding regions including upstream and downstream regions (Figure 2). Interestingly, methylation peaked in the upstream and downstream regions of genes and a sharp drop was observed at the boundaries of coding regions of all strains. Methylation within coding regions (gene body methylation) was observed mostly near the start and end of coding regions, gradually disappeared, and depleted at the center in all strains (Figure 3). The distribution of mC sites in different genomic features of other strains was similar to that of the wild-type. Compared to the wild-type, these increased proportions of mC sites were mainly found in exon regions. For example, compared to the wild-type showing 7.3% exon methylation sites, TdBCK1-S1 showed a marked increase to 15.1% and almost all the increased proportion of mC sites was found in near the start and end, but not the middle, of coding regions. Interestingly, when we measured the methylation levels around genic regions depending on the DNA context, hypomethylation was observed in the mutant strains compared to the wild-type strain. In addition, the methylation levels in exon regions of all strains were lower than any other genic features, i.e., upstream, downstream, and intron. Thus, although marked increases in the number of mC sites were mostly observed in the exon regions of the mutant strains, the methylation levels around genes of the mutant strains did not change significantly, i.e., still lower than those of the wild-type strain and the methylation levels in exon regions were lower than other genic regions.

Among all mC sites, 965,231 sites were common to all four strains (Figure 4A). Additionally, there were large numbers of mC sites that were common to two or three strain combinations. Furthermore, a number of strain-specific mC sites were identified; that is, 262,256, 512,497, 466,776, and 391,661 mC sites were specific for the wild-type, TdBCK1, TdBCK1-S1, and TcBCK1-S1, respectively, which represented 19, 30, 30, and 25% of the total, respectively. Among the common 965,231 mC sites, 84% were found in intergenic regions (Figure 4B). However, most of the strain-specific mC sites were found in genic regions; that is, 79, 77, 85, and 82% of the mC site specific for the wild-type, TdBCK1, TdBCK1-S1, and TcBCK1-S1, respectively, were in genic regions (Figure 4C). Within the genic regions, exons had the highest proportion (28–45%) of strain-specific mC sites (Figure 4D). It will be of interest to see how these strain-specific mC sites contribute to gene expression resulting in phenotypic changes.

In summary, compared with those of the wild-type strain, decreased global methylation levels and an increased number of mC sites were observed in the mutant strains. Methylation occurred in any DNA context, such as CG, CHG, and CHH, at a similar proportion in all four strains. The distribution of mC sites was not even, but clustered around intergenic and gene-poor regions across the contigs, with a notable exception of a domain in supercontig 4 of the TdBCK1 strain. In terms of genomic features, the distribution of mC sites demonstrated a preference for intergenic to genic regions. However, an increased distribution of mC sites in the genic regions was observed in the mutant strains, and these increased mC sites were mainly confined in exon regions. There were common mC sites among all four strains, but also strain-specific mC sites, which were mostly found in intergenic and genic regions, respectively. The methylation level of each mC site was not high, but showed a strong preference for ≤40% in all strains.

Quantification of the global DNA methylation level indicated the considerable changes in the DNA methylation of TdBCK1-S1. Therefore, we reasoned that the molecular mechanism of sectorization involves altered DNA methylation and investigated the expression profiles of the gene encoding DNA methyltransferase (DNMTase). A search for the DNMTase gene in the annotated C. parasitica draft genome sequence (Nuss et al., 2010) yielded three C5-DNA-methyltransferase genes, among which, one in scaffold_3:721439-722781 of the EP155 genome sequence assembly had a truncated C5-DNA-methylase domain. In addition, no corresponding RNA transcript was identified. Thus, considering it as a pseudogene, the other two genes designated as CpDmt1 (scaffold_2:1971293-1972939) and CpDmt2 (scaffold_6:348508-351624) were further analyzed for their functional contribution toward the DNA methylation mediated sectorization process. Based on the results of genomic sequence analysis, near full-length cDNA clones for CpDmt1 and CpDmt2 were obtained using RT-PCR with the primer pairs CpDmt1-mF1/CpDmt1-mR1 and CpDmt2-mF1/CpDmt2-mR1, respectively. Sequence comparison with the corresponding genomic sequences revealed that both CpDmt1 and CpDmt2 genes were intronless. The deduced CpDmt1 protein product (CpDMT1) consisted of 1,095 amino acids, with an estimated molecular mass of 121.1 kDa and pI of 5.63 (GenBank No. MF000328). The CpDmt2 gene consisted of two exons, with one intervening sequence of 30 bp. The deduced CpDmt2 protein product (CpDMT2) consisted of 1,259 amino acids, with an estimated molecular mass of 141.8 kDa and pI of 5.56 (GenBank no. MF000329). Phylogenetic analysis indicated that the cloned CpDmt1 and CpDmt2 genes, showing high similarity to repeat-induced point mutation (RIP) defective (RID) and defective in DNA methylation (DIM-2) of N. crassa, respectively, belonged to DNMT subfamilies of dnmt4 and dnmt1, respectively (Supplementary Figure S3 and Supplementary Table S7; Freitag et al., 2002; Ponger and Li, 2005).

Quantitative real-time RT-PCR was performed to obtain the expression profiles of CpDmt1 and CpDmt2. As shown in Figure 5, CpDmt1 (Figure 5A) and CpDmt2 (Figure 5B) expression levels were markedly down-regulated in the sectored progeny TdBCK1-S1. Additionally, the parental TdBCK1 and complemented TcBCK1-S1 strains showed significantly reduced (p < 0.01, one-way ANOVA) expression of both CpDmt1 and CpDmt2 compared to the wild-type. These results indicated that hypomethylation of mutant strains was correlated with the expression levels of both CpDmt1 and CpDmt2.

As the expression levels of both CpDmt1 and CpDmt2 were affected in TdBCK1-S1, we constructed CpDmt1- and CpDmt2-null mutants to examine the phenocopy of the sectorization. Eighty and 90 transformants for each CpDmt1- and CpDmt2-null mutant, respectively, were screened by PCR using a pair of outer gene-specific and inner hph primers (DMT1_EXT_F1&DMT1_EXT_R2 and DMT2_EXT_F1& DMT2_EXT_R2 in Supplementary Table S11) corresponding to -1,765 to -1,747 and 4,397 to 4,415 (relative to the start codon of CpDmt1) and -2,464 to -2,482 and 5,577 to 5,595 (relative to the start codon of CpDmt2), respectively. Transformants showing PCR amplicons of the expected size of 6.8 and 7.3 kb corresponding to the CpDmt1- and CpDmt2-null mutants, respectively, were obtained. These putative CpDmt1- and CpDmt2-null mutants were single-spored and single-spored progenies were further confirmed by Southern blotting analysis (Supplementary Figure S4).

The CpDmt1-null mutant showed retarded growth (Figure 6A) compared to the wild-type. However, compared to the CpBck1-null mutant showing severely retarded growth with thinner invasive feeding hyphae, the near-absence of the typical mycelial mat and aerial hyphae on the surface, and the absence of typical spore-forming structures, the CpDmt1-null mutant showed abundant sporulation, active mycelial growth and near-normal pigmentation (Kim et al., 2016). When the CpDmt1-mutant culture was aged, it started to show sectorization with robust mycelial growth (Figure 6B). The sectored progeny maintained the characteristics of active mycelial growth with reduced sporulation and restricted pigmentation. The sectored progeny differed from the wild-type in that fluffy mycelial growth was maintained and differentiation, such as pigmentation and sporulation, occurred but to a lesser extent than in the wild-type. In contrast, minor changes were observed in the CpDmt2-null mutant (Figure 6A). The CpDmt2-null mutant showed a similar growth rate to the wild-type. However, aerial hyphae around the colony margin disappeared with concurrent uncovering of pigmented spore-bearing structures, which differed from the wild-type showing the usual disappearance of aerial mycelia from the center (Figure 6A).

We, next examined DNA methylation of the CpDmt1- and CpDmt2-null mutants by measuring the changes in DNA methylation of representative genes using CyMATE. CyMATE indicated that both mutants showed reduced levels or near absence of DNA methylation (Supplementary Table S3). These results indicated that both genes are involved in DNA methylation. Two DNMTases, MoDIM-2 and MoRID, which are homologs of DIM-2 and RID, respectively, were observed in Magnaporthe oryzae (Jeon et al., 2015). Unlike the RID in N. crassa showing no evidence of DNMTase activity, both genes were suggested to be involved in DNA methylation, either alone or in combination. Moreover, it was interesting that the colony morphology of the MoDIM-2-null mutant in M. oryzae was fluffier than that of the wild-type. These results suggested that robust growth of mycelia, one of the signature characteristics of sectorization, is a common phenomenon governed by a set of genes controlled by changes in DNA methylation. However, changes in DNA methylation were specific depending on the DNMTases and resulting in different phenotypes as evidenced by the differences between CpDmt1- and CpDmt2-null mutants and other fungi.