M. oryzae wild-type strain KJ201 used in this study was obtained from the Center for Fungal Generic Resources (CFGR; hrtp://genebank.snu.ac.kr). All strains including wild-type strain and mutants constructed in this study were incubated on V8 juice agar medium (8% V8 juice [w/v], 1.5% agar powder [w/v], 10 N NaOH) or oatmeal agar medium (OMA, 5% oat meal [w/v], 2.5% agar powder [w/v]) at 25°C under constant fluorescent light. Mycelia used for RNA, DNA, and protein extraction were incubated in liquid complete media (LCM, 0.6% yeast extract [w/v], 0.6% casamino acid [w/v] and 1% sucrose [w/v]) at 25°C with 150 rpm shaking. Selection of hygromycin-resistant transformants was carried out using TB3 agar plates (0.3% yeast extract [w/v], 0.3% casamino acid [w/v], 1% glucose [w/v], 20% sucrose [w/v] and 0.8% agar powder [w/v]) supplemented with 200 ppm hygromycin B.

Nucleotide and protein sequences were procured and analyzed at Comparative Fungal Genetic Platform (CFGP; http://cfgp.snu.ac.kr) and National Center for Biotechnology Information (NCBI). Sequences of JmjC domain-containing protein were obtained from ChromDB database (www.chromdb.org). Analyses using Hidden Markov Model (HMM) profiling (http://hmmer.janelia.org) were performed to search for additional JmjC proteins in M. oryzae and to compare the protein sequence homology in other organisms. Protein sequences were aligned using ClustalW and T-coffee algorithm. Phylogenetic tree was constructed with MEGA 5.2 using neighbor-joining method with a Poisson correction model and a bootstrap of 1,000 replicates. Information about domain architectures was obtained from InterProScan.

Fungal genomic DNA was extracted by two different methods according to the purpose. For large-scale screening of transformants, fungal DNA was extracted from mycelia on TB3 agar plates using quick extraction protocol (Chi et al., 2009). For Southern blot analysis genomic DNA was extracted using phenol-chloroform method. Total RNA was extracted with the Easy-Spin total RNA extraction kit (iNtRON Biotechnology, Seoul, Korea) according to the manufacturer’s instruction. cDNA for expression analysis by quantitative real-time PCR (qRT-PCR) was synthesized by using ImProm-II Reverse Transcription System kit (Promega, Madison, WI, USA) following the manufacturer’s instruction.

To construct gene deletion mutant, homologous recombination strategy was used. The original sequence of MGG_04878.7 was replaced with a knock-out construct containing 1.4 kb of hygromycin B phosphotransferase (HPH) cassette fused with ~1.2 kb 5′- and 3′-flanking regions by double joint PCR. Protoplast isolation and fungal transformation were carried out using PEG mediated transformation. Transformants resistant to hygromycin B within TB3 media were selected. Disruption of genes was verified by PCR using primers for screening (Supplementary Table 1). Southern blot analysis using genomic DNA was performed to confirm correct integration of HPH cassette. Genomic DNA fragments, digested with restriction enzyme, were separated by agarose gel electrophoresis and transferred to nitrocellulose membrane, Hybond-N+ (GE Healthcare, Piscataway, NJ, USA). The blot was hybridized with probe labeled with P32 by using Rediprime II Random Prime Labeling System kit (Amersham Pharmacia Biotech, Piscataway, NJ, USA) and the hybridized membrane blot was exposed to imaging plate, BAS-2040 (Fuji Photo Film, Tokyo, Japan) and read by phosphorimage analyzer, BAS-1000 (Fuji Photo Film).

Complementation strain for MoJMJ1 was generated by the following procedures. DNA fragment containing ORF and upstream sequences were amplified from KJ201 genomic DNA. The resulting PCR product and pII99 plasmid containing geneticin-resistant gene as a selection marker were co-transformed into protoplast of deletion mutant ΔMojmj1. Complementation strains were selected on geneticin-supplemented media and verified by PCR using primer pairs ORF_F_screening and ORF_ R_screening (Supplementary Table 1) for existence of the whole MoJMJ1 ORF, which was absent in the deletion mutant.

Mycelial growth was measured as colony diameter of 9-day-old culture on modified complete agar media (1% glucose [w/v], 0.2% peptone [w/v], 0.1% yeast extract [w/v], 0.1% casamino acid [w/v], 0.1% trace element [v/v], 0.6% sodium nitrate [w/v], 0.05% potassium chloride [w/v], 0.05% magnesium sulfate [w/v], 0.15% potassium dihydrogen phosphate [w/v], 1.5% agar powder [w/v]) and 4 mm diameter inoculums were taken from 5-day-old minimal agar media in marginal region.

Conidiation was measured by counting the number of asexual spores in 15 μl of conidia suspension on a haemacytometer. For conidiation assay, conidia were harvested from 7-day-old V8 juice agar media using 5 ml of sterilized distilled water. Conidiophore development was monitored by observing agar block (~1 × 1 cm) from 9-day-old culture following scrape-off.

For measuring the rates of conidial germination and appressorium formation, conidia were harvested from 7-day-old V8 juice agar media with sterilized distilled water and filtered with two-layers of miracloth (Calbiochem, San Diego, CA, USA). Thirty microliters of conidial suspensions (3 × 10⁴ conidia/ml) was dropped on hydrophobic coverslips (W. Knittel Glass, Braunschweig, Germany) with three repeats. The droplets were incubated in moist-lasted airtight containers at room temperature. Samples were incubated for 12 and 8 h for measuring conidial germination and appressorium formation/development rates respectively. For nikkomycin Z sensitivity test, conidia from wild-type, deletion mutant, and complementation strain were incubated with or without 100 μM of nikkomycin Z, and examined after 2.5 h.

For pathogenicity tests and wound inoculation, conidia formed on 7-day-old V8 juice agar media were rubbed using sterilized cotton swabs dipped in distilled water containing Tween 20 (250 ppm) and filtered with two-layers of autoclaved miracloth. For sheath assay, the abovementioned collection procedure was followed, except that Tween 20 was not used. Final concentration of spore suspensions was maintained at 5 × 10⁴ conidia/ml for pathogenicity test and wound inoculation, and 2 × 10⁴ conidia/ml for sheath assay. Three- or four-week-old rice seedlings of Oryza sativa cv. Nakdongbyeo were used for above tests.

For pathogenicity test, spore suspension (10 ml of total volume) was sprayed onto 3-week-old rice seedlings which was then placed in a dark, humid dew chamber for 18 h at 25°C. Next, the seedlings were transferred to a growth chamber maintained at 80% humidity and 25°C. Leaves of the rice seedling were collected on day 7 after the inoculation and scanned.

Leaves of Nakdongbyeo were used for wound inoculation. Conidial suspensions (5 × 10⁴ conidia/ml) were dropped at wounds, which were picked about 10 times by tip of a sterilized pin. Inoculated leaves were incubated in a humid container at 25°C. Lesions were observed and scanned at 6 days post inoculation (dpi).

Sheathes of Nakdongbyeo in 5-6 leaf stage were used for rice sheath assay. Rice sheathes were inoculated with conidial suspension (3 × 10⁴ conidia/ml) and incubated in a moist container at 25°C. The sheath samples were sliced by razor blade and observed with light microscope.

Frozen fungal mycelia were homogenized by homogenizer and added to PRO-PREP protein extraction solution (iNtRON Biotechnology). The homogenates were centrifuged (13,000 rpm, 10 min, 4°C) and the supernatant was collected. Equal amounts of homogenized protein (15 μg) were applied to 12% SDS-PAGE gel (Bio-Rad, Hercules, CA, USA), and electro-blotted onto an Immunobilon-P PVDF membrane (Bio-Rad). For the gel running and transfer procedures, ReliaBLOT^® running buffer (20×) and transfer buffer (10×) (Bethyl, Montgomery, TX, USA) were used, respectively. The blots were probed with polyclonal histone H3 and H3K4me3 antibody (Active Motif, Carlsbad, CA, USA), using Pierce^® Fast Western Blot Kit, ECL Substrate (Bio-Rad).

As a first step toward elucidating importance of histone demethylation in M. oryzae, we searched the proteome for putative HDM in M. oryzae using hidden Markov model built on known HDM sequences, and found that there are a total of seven genes encoding putative HDM among eleven proteins containing JmjC-domain. The following phylogenetic analysis showed where the seven genes could be placed in reference to known JmjC protein sequences of human, rice, Arabidopsis and yeast. Among the predicted seven genes in M. oryzae, MGG_04878 was predicted to belong to JARID group; MGG_09186 belonged to JMJD2; MGG_01543 and MGG_04401 to JmjC domain only groups, and MGG_09841, MGG_02045 and MGG_01068 appeared to be fungal specific through phylogenetic analysis (Supplementary Fig. 1). Among these genes encoding putative JmjC domain-containing proteins, we focused on a gene encoding JARID group histone lysine demethylase (MGG_04878) for further analysis, since we were able to obtain deletion mutants for the gene, and designated it as MoJMJ1 (M. oryzae jumonji 1).

Several orthologs MoJMJ1 have been characterized in species as diverse as human, Drosophila melanogaster, Caenorhabditis elegans to Saccharomyces cerevisiae and Schizosaccharomyces pombe. From these studies, it was shown that majority of the JARID1 group proteins share the same target histone residues (di- and tri-methyl H3K4) and contain highly conserved cofactor-binding sites required for histone lysine demethylase activity (Agger et al., 2008; Huarte et al., 2007; Klose et al., 2006; Secombe et al., 2007). When we compared domain architecture, we found that MoJMJ1 protein was predicted to have JmjN (IPR003349), JmjC (IPR003347), BRIGHT/ARID (IPR001606), C5HC2-zf (IPR004198), and PHD domain, all of which are typical domains of human JARID1 orthologs (Fig. 1A). Multiple sequence alignment shows that all the key residues for cofactor binding is conserved in MoJMJ1 (Fig. 1B).

To investigate the role(s) of a JARID group demethylase in fungal development and pathogenicity, we generated deletion mutants in which the native MoJMJ1 sequence in wild-type strain, KJ201, was replaced by HPH cassette (Fig. 2A). Correct gene replacement event was verified through PCR screening and following Southern blot analysis in the mutants (Fig. 2B). Absence of transcript from Mo-JMJ1 locus in ΔMojmj1 mutant strain was confirmed by RT-PCR (Fig. 2C).

Based on similarity at primary amino acid sequences, we predicted that MoJMJ1 is a histone lysine demethylase targeting H3K4me3, which serves as a mark for active transcription (Kooistra and Helin, 2012; Lee et al., 2007b). To test if deletion of MoJMJ1 affects global H3K4me3 level, we performed western blot analysis using antibody raised against H3K4me3. The western blot analysis showed that global H3K4me3 level increased by at least 70% in the mutant compared to the wild-type strain, suggesting that MoJMJ1 possesses HDM activity for H3K4me3 (Fig. 2D, Supplementary Fig. 2).

Phenotypes of ΔMojmj1 were examined to investigate the roles of Mo-JMJ1 in vegetative and developmental stages. First, effects of MoJMJ1 deletion on fungal vegetative stage were monitored. The diameter of ΔMojmj1 mutants showed reduction of 20%, compared to wild-type at 9 dpi on complete agar media (Table 1). In addition, autolysis was observed in both deletion mutants (Fig. 3A). These observations indicate that MoJMJ1 is important for normal vegetative growth in M. oryzae.

When we compared asexual reproduction between the wild-type and mutant by measuring the ability to produce conidia on axenic culture, we found that concentrations of conidial suspension obtained from the colony of ΔMojmj1 mutants were 25-40% of those of wild-type, indicating that MoJMJ1 is involved in asexual reproduction as well (Table 1). To further investigate the reason why asexual reproduction of the mutant was reduced in the mutant, we monitored conidiophore development on agar blocks. Although it appeared that the mutant was able to form at least three conidia in a sympodial pattern at a conidiophore, overall number of conidiophores decreased in ΔMojmj1 mutants (Fig. 3B). This suggests that MoJMJ1 may be associated with efficient development of conidiophore from hyphal tissue.

Normal development of appressoria is a pivotal step for host infection in M. oryzae. To investigate whether MoJMJ1 affects conidial germination and appressorium formation, a drop of conidial suspension was incubated on an artificial hydrophobic surface (coverslip). Most of conidia germinated in both the wild-type and deletion mutants (Table 1). Unlike the wild-type conidia developing a single appressorium at a germ tube tip, conidia of ΔMojmj1 showed bipolar germ-tubes that started to branch at 8 hours post incubation (Fig. 3C, left panels), and 95% of the conidia in ΔMojmj1 failed to develop appressoria, suggesting requirement of MoJMJ1 in appressorium formation (Table 1). Interestingly, abnormal appressorium started to develop at the tip of excessively elongated germ-tubes of ΔMojmj1 at around 20 hours post inoculation (hpi) and proportion of the mutant conidia that developed abnormal appressoria increased up to 30-40% at 36 hpi on coverslips (Supplementary Fig. 3). Such delay in appressorium formation prompted us to investigate the implication of MoJMJ1 in regulating signaling pathways involved in appressorium formation. To this end, we employed chemical complementation using cAMP, CaCl₂, and 1,16-hexadecanediol. However, adding those chemicals separately or in combination into conidial suspension of ΔMojmj1 on hydrophobic surface was not able to complement the defects in appressorium formation of the mutants. It was previously shown that M. oryzae could develop appressoria not only at the germ tube but also at hyphal tips (Kim et al., 2009). When we tested the effect of MoJMJ1 deletion on the formation of hyphal-driven appressoria, we found that the mutant lost the ability to develop appressoria at the hyphal tip as well (Fig. 3C, right panels). These results suggest that MoJMJ1 is essential for normal development of appressorium from two different types of cells, germ-tubes and hyphae. When we ectopically introduced MoJMJ1 gene plus 1.2 kb putative promoter sequences back to the deletion mutant, all the defects in vegetative growth, conidiation, and appressorium formation were restored to the wild-type level (Fig. 3, Table 1).

Next, we tested whether MoJMJ1 is indispensable for fungal pathogenesis, given the inability of the mutant to develop appressoria on artificial hydrophobic surface. To evaluate pathogenicity of the mutant, we spray-inoculated the rice plants with conidial suspensions of wild-type and ΔMojmj1 strain, and compared the symptom development at 5 dpi. While the leaves challenged with wild-type strain showed typical spindle-shaped lesions, leaves inoculated with the deletion mutants remained asymptomatic (Fig. 4A). To our surprise, the complementation strain MoJMJ1c, however, was not able to cause disease at all on the rice plants (Fig. 4A). Such inability of complementation strains to cause disease was consistently observed in other complementation strains that were generated by a number of independent experiments. Upon failure of complementation for pathogenicity defect, we examined the mutant and complementation strain in detail to elucidate the nature of complementation failure by monitoring infection processes at the single cell level using sheath assay method. Unlike wild-type strain penetrating host cells via appressorium at 24 hpi and colonizing neighboring cells at 48 hpi, most of the conidia from deletion mutant was not able to form appressoria, which is consistent with our observation for in vitro appressorium formation of the mutant. Mishap appressoria, which developed at the tip of abnormally elongated germ tube of the mutant, appeared to have lost the ability to penetrate underlying host cells. Similarly, MoJMJ1c appressoria did not show any sign of host cell penetration even at 48 hpi, although the complementation strain developed appressoria that are indistinguishable from those of wild-type (Fig. 4B). One possibility of such penetration defect is disruption of cell wall integrity. We tested this possibility using Nikkomycin Z, an inhibitor of cell wall biosynthesis. While wild-type conidia were able to develop appressoria when treated with 10 ng/ml of Nikkomycin Z, both ΔMojmj1 and MoJMJ1c conidia exhibited ballooning of the germ tube tip and were not able to develop appressoria at the same concentration of Nikkomycin Z, suggesting defective cell wall in the mutant and complementation strains (Supplementary Fig. 4A). We then went further to test effect of such defect in cell wall on generation of turgor pressure within appressorium using cytorrhysis assay. The assay result showed that appressoria of both the mutant and complementation strains tend to collapse in less frequency at 3 M glycerol than the wild-type, indicating problem(s) in cell wall integrity of appressoria, which is important for appressorium functionality (Supplementary Fig. 4B).

We then ask if the mutant and complementation strain retain the ability to grow inside plant tissues by allowing direct entry of the fungus into host tissues through a wound site without appressoirum-mediated penetration process. To test this, the conidial suspension of ΔMojmj1 or MoJMJ1c was injected into plant tissues using a 22-guage syringe. In our assay, the wild-type strain developed large disease lesions but MoJMJ1c as well as ΔMojmj1 strains were unable to develop disease lesions. These data indicate that introduction of MoJMJ1 gene plus 1.2 kb upstream sequences into ΔMojmj1 strain can complement pre-penetration stage developments, but not penetration and post-penetration development.

For partial complementation of phenotypes, we first considered the possibility that unsolicited mutations occur during preparation and/or transformation of protoplast. To check this, we evaluated pathogenicity of two independent ectopic mutant strains (MoJMJ1e_52 and MoJMJ1e_185) that are obtained from the same batch of transformants as the deletion mutant strain. The ectopic mutants harbor knockout construct with intact MoJMJ1 gene. Therefore, if there is additional mutation(s) during protoplast preparation or transformation of protoplasts with knockout construct, they will share the mutation(s) and will be non-pathogenic as well. Spray-inoculation with these strains clearly showed that ectopic mutants are fully pathogenic on the rice plants, allowing us to rule out the possibility of presence of unsolicited mutation(s).

The second possibility that we considered is the length of 5′ upstream region required for complete complementation. To explore this, we extended 5′ upstream region of the gene included in a new complementation construct up to 2.1 kb. When we introduced the new complementation construct into ΔMojmj1 and screened for geneticin-resistance, we found that at least 80% of the resulting transformants were able to form appressoria on plastic coverslip (Fig. 5A). Among them, we randomly chose three transformants (MoJMJ1c_2-1, MoJMJ1c_2-2, and MoJMJ1c_2-3) and carried out spray-inoculation on rice plants using spores from the three strains as inoculum. Our data showed that none of them are pathogenic, suggesting that failure to complement pathogencity defect of the deletion mutant is unlikely to be associated with promoter length (Fig. 5B).

Lastly, we checked out expressions of two genes using qRT-PCR: (1) MoJMJ1 gene that is ectopically introduced into the complementation strain (construct having 1.2 kb promoter sequences) and (2) CON7 gene, which is known to be related to Nikkomycin Z sensitivity (Odenbach et al., 2007). Interestingly, MoJMJ1 gene was shown to be up-regulated in the complementation strain, while expression of CON7 was down-regulated in the mutant and restored to the wild-type level in the complementation strain (Fig. 6).