The Nep1-like protein family of Magnaporthe oryzae is dispensable for the infection of rice plants

The necrosis- and ethylene-inducing protein 1 (Nep1)-like proteins (NLPs) are a class of microbe-associated molecular patterns widely distributed across diverse groups of plant-associated microorganisms. In spite of the cytotoxic activity in dicot plants, the role of most NLPs in the virulence of plant pathogens is still largely unknown. We showed that the MoNLP family of rice blast fungus varied very little in amino acid sequence, transient expression of three MoNLPs induced cell death and the production of reactive oxygen species in Nicotiana benthamiana, and the expression of MoNLPs was induced during infection of susceptible rice plants. To further investigate the biological role of the MoNLP family, a marker-free gene replacement vector was developed and used to knock out the whole family in Magnaporthe oryzae. Results showed no significant difference in disease levels caused by wild type and the quadruple ΔMoNLP mutant strains. Likewise, the sporulation and radial growth of the two strains were similar under various unfavorable cultural conditions including malnutrition and abiotic stresses. These observations demonstrated that the MoNLP family is dispensable for the fungal tolerance to the tested adverse cultural conditions, and more importantly, for the virulence of blast fungus on susceptible rice plants.

the pathogen under these non-cytotoxic conditions is still largely unknown. Motteram et al. (2009) have demonstrated that the only NLP gene of M. graminicola is dispensable for the fungal pathogen to cause disease on wheat plants but the role of a multi-membered NLP family in host colonization has not been reported.
Here, we describe our study on the potential biological role of NLP family in Magnaporthe oryzae, the causal agent of rice blast disease, which has four MoNLPs genes in the genome 7,17 . Through a gain-of-function genetic screening, we initially identified a MoNLP gene that encoded an elicitor triggering typical immune responses in Nicotiana benthamiana. Subsequently, we investigated several aspects of the MoNLP family of the fungus including the amino acid sequence variation among lab strains and field isolates, expression profiles during infection, and the cytotoxicity of each member of the protein family. More importantly, we developed a pop-in/pop-out gene replacement vector and successfully deleted all four MoNLP genes from the genome of M. oryzae. The potential roles of MoNLPs in fungal virulence and tolerance to various adverse cultural conditions were further evaluated.

Results
Activation tagging identified an MoNLP gene from M. oryzae that elicited cell death when overexpressed in N. benthamiana plants. To Supplementary  Table S1). Sequencing of the insert ends of G7, a cell death-inducing clone, revealed that the insert contained two annotated fungal genes MGG_10531 and MGG_10532 (see Supplementary Fig. S2A). To further determine which of these two genes was responsible for the cell death, restriction endonucleases that specifically disrupted the individual gene were used to truncate the G7 plasmid clone, and the transient expression assay showed that MGG_10532, which encodes the MoNLP4 17 , was both necessary and sufficient for the G7 induced cell death in N. benthamiana plants (see Supplementary Fig. S2B). Genes responsible for the elicitor activity of other three clones were subsequently identified with a similar approach, and they were predicted to encode two hypothetical proteins and a Ras-2 protein, respectively (see Supplementary Table S1). Among these candidate elicitors, MoNLP proteins have the predicted signal peptide for protein secretion, and the function of NLPs in the microorganism is still obscure; we therefore chose the MoNLP family for further investigation.
The MoNLP family is highly conserved across strains of M. oryzae. Multiple studies have revealed that NLP family members can vary significantly in number among different microbial species (see Supplementary  Table S2) although they share the highly conserved domain 7 , indicating the potential differentiation in NLP function associated with distinct microorganisms. Previous studies have identified four NLP-encoding genes in the blast fungus: MGG_08454 (MoNLP1), MGG_00401 (MoNLP2), MGG_02332 (MoNLP3) and MGG_10532 (MoNLP4) 7,17 . We investigated the sequence variation of MoNLPs among three sequenced M. oryzae strains 20 and a panel of field isolates collected from different geographic regions of China. The results showed that the MoNLP family rarely varied: MoNLP1, MoNLP2, and MoNLP4 were individually identical, and MoNLP3 shared 99.6 ~ 100% of identity across the 21 tested strains and isolates (see Supplementary Fig. S3), implying that MoNLPs may have important roles in the biology of M. oryzae.

Expression of MoNLPs triggered oxidative burst and cell death in N. benthamiana. Several
NLPs have been shown to be able to induce cell death preferentially in dicotyledonous plants 9,12,14 . To investigate the MoNLP-induced cellular responses on N. benthamiana plants, cDNAs of individual MoNLP genes were cloned and engineered downstream of the inducible promoter of a modified vector of pER8 21 , and the resulting constructs were used for Agrobacterium-mediated transient expression assay. Tissue staining of Agrobacterium-infiltrated leaves showed the production of reactive oxygen species (ROS) and the cell death of plant cells at 24 hours after treatment with the chemical inducer for constructs expressing MoNLP1, MoNLP2 and MoNLP4, but not for MoNLP3 and the GFP control ( Fig. 1), indicating that MoNLPs differed in cytotoxic activity on N. benthamiana.

Members of the MoNLP family were induced during infection of rice plant.
To investigate the potential role of MoNLPs during the interaction between M. oryzae and its host plant, we inoculated susceptible rice seedlings with wild type blast fungus strain P131, and analyzed levels of MoNLP transcripts throughout the infection process by quantitative real-time PCR. Results showed that all four members of the MoNLP family were up-regulated during the infection process but with seemingly distinct induction profiles. MoNLP2 was strongly induced at 8 hours post inoculation (hpi) but subsided quickly afterward. Transcript levels of MoNLP4 were significantly up-regulated at multiple time points in both early and late stage of the infection. The up-regulation of MoNLP1 initiated from 48 hpi, when the symptom of chlorosis started to develop, whereas a significant transient induction of MoNLP3 could be observed at a rather late stage of infection (96 hpi), when infected leaf tissue had already collapsed and fungal sporulation was visible on the abaxial surface of the inoculated leaf (Fig. 2). The diversified patterns of transcripts accumulation indicated that MoNLPs were likely involved in both biotrophic and necrotrophic phases of the infection.
Knocking out the MoNLP family through pop-in/pop-out gene replacement. To further elucidate whether the infection-associated activation of MoNLPs contributed to the virulence of blast fungus, we attempted to knock out all four members of the MoNLP family from the genome of M. oryzae. To perform the gene replacement, we firstly constructed a plasmid vector, pMFKO-DONR, which contained a hygromycin B resistance gene for positive selection, a conditional lethal HSVtk gene 22 for negative selection, and the attP cassette of a pDONR vector (Gateway technology, Invitrogen) for cloning of DNA fragments required for in situ homologous recombination (Fig. 3). Subsequently, upstream and downstream DNA fragments of individual MoNLP coding regions were amplified and fused together by PCR and recombined with the vector by BP recombinase to produce pMFKO-NLP. These plasmids were used to sequentially delete all four MoNLP genes from the genome of P131 strain (Fig. 4). Interestingly, while we were constructing the quadruple mutant strainΔMoNLP 1-4 , we found that knocking out either MoNLP1 or MoNLP4 had little impact on the expression of MoNLP3 during infection but simultaneous disruption of both genes strongly enhanced transcript levels of MoNLP3 at 96 hpi (Fig. 5). This synergistic effect on gene expression between MoNLPs indicated that certain levels of redundancy may exist between members of the MoNLP family during disease.
The MoNLP family is dispensable for the infection of rice plants and the growth under various stress conditions. Susceptible rice seedlings were spray-inoculated with the quadruple mutant and wild type P131 strains to investigate the potential role of MoNLP genes in fungal virulence. Five days after inoculation, similar levels of disease symptom in terms of the size and density of lesions were observed on plant leaves for both treatments (Fig. 6A). Likewise, the biomass of the quadruple mutant and wild type strains in the inoculated leaves hardly differed as indicated by the quantitative PCR analysis of the fungal DNA in diseased samples (Fig. 6B). We also inoculated detached rice leaves with conidial droplets and observed no significant difference in disease symptom and fungal biomass between quadruple mutants and wild type strains (see Supplementary Fig. S4). Moreover, we examined the growth of infectious hyphae around infection sites on the leaf sheath of rice plants. Penetration sites of single spores were classified into three categories based on the number of plant cells (1, 2-9 and > 10 cells) that the infectious hyphae had colonized. At 72 hpi, the percentages of each category of the infection sites were similar between wild type P131 and the quadruple mutant strain (Fig. 6C), indicating that deletion of the MoNLP gene family did not affect the fungal growth on leaf sheath as well. These results collectively demonstrated that the MoNLP gene family is not required for M. oryzae to infect the rice plant.
To determine whether the MoNLP family was required for the growth of M. oryzae, wild type and the quadruple mutant strains were grown on CM medium, 1.2% agar, Czapek-Dox medium, and nitrogen or carbon starvation medium. Results showed that the radial growth of the quadruple mutant was similar to that of wild type strain on these media (see Supplementary Fig. S5A). In addition, both strains produced similar levels of conidia when cultured on the oatmeal-tomato agar (OTA) medium (see Supplementary Fig. S5B). These findings suggested that the MoNLP family was not central to the growth and sporulation of M. oryzae under these conditions.
To examine the role of MoNLP family in the fungal response to diverse stress conditions, wild type and the quadruple mutant strains were grown on rich medium supplemented with SDS, congo red, 1.2 M sorbitol, and the iron chelator diethylenetriamine pentaacetic acid (DTPA), respectively. Radial growth of the quadruple mutant was found similar to that of the wild type strain (see Supplementary Fig. S6A), indicating that disruption of the MoNLP family did not compromise the integrity of fungal cell wall and membrane, and the tolerance to osmotic stress or iron limitation. Likewise, the growth of the two strains showed no difference under diverse pH (5.0 and

Discussion
Current understanding of the biological role of NLPs during disease process is very limited, although genes encoding this class of proteins have been found in a wide range of microorganisms. Previous studies have shown that four genes encoding NLP exist in the genome of M. oryzae 7,17 . Our analysis revealed a remarkable conservation of MoNLPs among all the tested field isolates and lab strains (see Supplementary Fig. S3), implying that these genes may be essential for the survival of M. oryzae in nature. Interestingly, we observed that transcript levels of all four MoNLPs were induced in either or both of the biotrophic and necrotrophic phases during the infection of susceptible rice plants (Fig. 2). The up-regulation of NLP gene expression during infection has been described in other plant-pathogen interactions as well. In necrotrophic fungal pathogen Botrytis cinerea, the induction of BcNep1 and BcNep2 is at early and late stage of infection, respectively 23 . In addition, transcript levels of most of the expressed PsNLPs in P. sojae 12,24 and two VdNLPs in V. dahliae 13 are highly induced at the late stage of infection; in contrast, the induction of four HaNLPs in the obligate biotroph oomycete Hyaloperonospora arabidopsidis occurs at the early stage of infection 16 . Notably, in hemibiotroph M. graminicola, the expression of the only NLP gene is prominently enhanced during a specific period near the end of the symptomless stage of the infection process 11 . These observations indicate that NLPs are naturally involved in the interactions between microbial pathogens and their host plants, but the complexity of gene induction profiles implies that these proteins may have diversified roles during infections.
NLPs are often regarded as virulence factors of pathogens on dicots due to the cytotoxic activity 25,26 . For instance, ectopic expression of an NLP gene from F. oxysporum in Colletotrichum coccdes, a fungal pathogen used for biological control of weeds, strongly enhances the virulence of C. coccdes on the weed Abutilon theophrasti 27 ; silencing several PcNLPs in P. capsici leads to reduced virulence on pepper leaves 15 . However, the impact of some cytotoxic NLPs on pathogen virulence on dicots appears to be host dependent. For example, VdNLP1 and VdNLP2 are two genes encoding cytotoxic NLPs from V. dahliae, and disruption of either VdNLP compromises the fungal virulence on tomato and Arabidopsis, whereas, on N. benthamiana plants, the disruption of only VdNLP1 but not VdNLP2 affects the virulence 13 ; the fungal virulence is not affected on cotton plants even when both genes were disrupted 28 . Likewise, the EccNip from Erwinia carotovora subsp. carotovora is required for full virulence of bacteria on potato tuber but not in potato stem or on other tested host plants 14 . Moreover, other studies show that deletion of a member of NLP family in B. cinerea or F. oxysporum does not impair the fungal virulence on tomato, N. benthamiana 23 , or coca plants 29 , suggesting that NLPs may be dispensable for the virulence of pathogens. It is also possible that in these studies additional NLP genes may compensate the loss of a member of the family, which may mask the potential phenotype of mutant strains.
Gene family expansion is a common phenomenon in nature that leads to gene redundancy and functional diversification, which have also been reported for NLPs 12,13,16,30 . In our study, the expression of MoNLP3 was significantly upregulated at 96 hpi in the ΔMoNLP1/ΔMoNLP4 double mutant compared to single knock-out or wild type strains (Fig. 5), implying that MoNLPs may be functionally complementary during infection of rice plants. To eliminate the potential redundancy between family members, we therefore, for the first time, disrupted all four MoNLP genes in the blast fungus, and results showed that the loss of entire multi-membered MoNLP family did not compromise the fungal virulence on rice (Fig. 6). This is unexpected especially considering the extensive upregulation of MoNLPs' expression during the infection process. However, it cannot be ruled out that other components or systems in M. oryzae may compensate the impact of the loss of MoNLP family on fungal  virulence, and mask the phenotype of the quadruple mutant; alternatively, the enhanced MoNLPs' expression may be responsive to unknown stimuli generated during the infection process rather than boosting the fungal virulence per se. Previous studies also show that deletion of the only MgNLP from M. graminicola does not affect fungal virulence on wheat 11 . These results collectively demonstrate that the NLP family are not essential for fungal infection of monocot plants.
Results of previous research show that NLPs from monocot-colonizing pathogens do not trigger necrosis on monocots 11,18 , NLP genes are activated in obligate biotrophic pathogens during infection 16 , and notably, insect pathogenic fungi, non-pathogens, and other animal-related fungi also carry NLP genes in their genomes 7,10 . Hence, it is reasonable to assume that NLPs may have essential roles independent of phytotoxicity during certain fundamental biological processes. It has been reported that VdNLP1 is required for the vegetative growth and production of conidiospore in V. dahliae 13 , and several VdNLPs are induced at a later stage of cultivation in liquid Czapek-Dox medium or by supplementing the medium with cotton root 28 . However, vegetative growth and sporulation were not affected in the quadruple mutant of M. oryzae in this study (see Supplementary Fig. S4). Many NLPs have a signal peptide and are secreted outside of the cells 8,9 , and evidence also shows that the expression of SsNep2 in S. sclerotiorum is upregulated by the physical properties of the contact surface 31 . Thus, the rich variation in NLP expression profiles may reflect the complexity of environmental cues encountered by microbes during their life cycle. Further investigation of the nature of these cues may provide helpful clues for dissecting the biological roles of MoNLPs.
In this study, we developed the vector pMFKO-DONR to delete four MoNLP genes in M. oryzae via the pop-in/pop-out method. This marker-free approach is commonly used for homologous replacement of genes in bacteria 32 , and is especially convenient when multiple genes are targeted for deletion. The efficiency of homologous recombination depends on where the target gene is located in the genome, and it has been reported that the average rate of targeted gene replacement is only about 7% in M. grisea 33 . Our results demonstrated that this approach works well in the blast fungus. Thus, the vector pMFKO-DONR and related experimental protocols may be useful for reverse genetic study in other haploid fungi where potential gene redundancy is involved as well.

Materials and Methods
Isolation of genomic DNA. For PCR screening of in situ transformants, the genomic DNA of M. oryzae was isolated from mycelia grown in liquid CM medium for 3 days according to Cenis 34 . To construct the genomic library, mycelia were harvested and ground in liquid nitrogen, and DNA was extracted with cetyltrimethylammonium bromide (CTAB) according to the protocols 35,36 .   Histochemical staining. Trypan blue staining was performed as described 37 . For DAB staining, samples were incubated in the DAB staining solution (1 mg ml −1 , pH 3.8) overnight, and destained 5-10 min subsequently in boiling ethanol 38 .

Construction of
The inoculated sheath was fixed using formaldehyde (ethanol-formaldehyde-acetic acid, 80:3.5:5, by vol.) as described 39 . The plant material was stained with lactophenol cotton blue for 6 hours at room temperature 40 . Vector construction. To construct the vector pCBER DEST used for inducible expression of MoNLP genes, the attR cassette of a Destination vector compatible with the GATEWAY ™ cloning technology was cloned into XhoI and SpeI sites of the target expression cassette of the pER8 vector 21 . This modified inducible transcription unit and the G10-90-XVE cassette of pER8 were amplified by PCR, and cloned into the HindIII and XbaI (blunted) sites, respectively, of pCB302 41 to obtain pCBER DEST.
To construct pMFKO-DONR, the attP cassette of pDONR201 (Invitrogen, USA) was amplified by PCR and cloned into the pGEM-T Easy vector (Promega, USA) carrying the hygromycin B phosphotransferase gene. The resulting plasmid was linearized by PCR with primer pair pTGD For and pTGD Rev (see Supplementary  Table S3), and further ligated with the negative selection marker HSVtk gene that was amplified from the plasmid pHH-5 42 to obtain pMFKO-DONR (Fig. 3).
To disrupt MGG_10531 in the G7 plasmid, SmaI and HindIII were used to digest the G7 plasmid, and the resulting fragment was blunted and self-ligated to obtain the plasmid clone ΔMGG_10531. Likewise, ClaI and PstI were used to digest G7 plasmid to disrupt MGG_10532, and the resulting fragment was blunted and self-ligated to obtain the plasmid cloneΔMGG_10532.
Targeted disruption of four MoNLP genes. Genomic DNA fragments over 1 kb in length flanking the coding region of the targeted MoNLP gene were amplified and fused together by PCR with gene-specific primers bearing attB sites at 5′-ends (see Supplementary Table S3). This DNA product was introduced into the pMFKO-DONR by BP recombinase to produce the vector pMFKO-NLP, which was used for the PEG/CaCl 2 mediated transformation of the protoplasts of blast fungus 43,44 . The transformed protoplasts were selected on 1% top agar supplemented with 250 μg ml −1 hygromycin B, and the resulting transformants were further screened for in situ homologous recombination by PCR analysis. Colonies bearing targeted integrations were transferred to OTA medium containing 5 μM 5-fuoro-2′-deoxyuridine (F2dU) to enrich cells having lost the counter-selection marker HSVtk gene. Conidia were harvested and grown on CM medium supplemented with 100 μM F2dU, and resistant candidates were subjected to PCR analysis to obtain mutants with the deletion of targeted MoNLP gene. The primers used for knocking out MoNLP genes are listed in Supplementary Table S3.
Gene expression analysis. Total RNA was extracted from about 100 mg inoculated leaves by Trizol (Generay), and cDNA was synthesized by M-MLV (Takara). Transcript levels of MoNLP family were determined by quantitative real-time RT-PCR (Takara). The genes encoding Actin and 40 S ribosomal protein S3aE of M. oryzae were used as reference genes. Each sample has three technical repeats. The experiments were repeated at least twice. Primer pairs used in this study were listed in Supplementary Table S3.
Plate assay of fungal growth, sporulation and stress tolerance. Radial growth of fungal mycelia was determined by placing 5 μl conidial suspension (1 × 10 4 conidia ml −1 in water) on the center of an agar plate supplemented with various ingredients 45 . The inoculated plates were incubated at 28 °C in darkness. The diameter of colonies was measured at 6 days. The media used for assays include the CM medium 46 , CM medium supplemented with 1.2 M sorbitol or 200 μg/ml congo red 47 , 1.2% agar, Czapek-Dox medium, nitrogen or carbon starvation medium 48 and rice medium with pH 8.0 or 5.0 49 , 1/10 CM medium supplemented with 8 μM DTPA 50 .
For quantification of sporulation, M. oryzae strains were cultured on OTA for 12 days 51 . Each strain has three repeats. Four discs (ϕ5 mm) harvested about 1 cm from the edge of plate were put into 2 ml tube, and 0.5 ml sterile water was added. Then, each sample was shaken vigorously for 1 minute. The conidial suspension was counted by haemocytometer.

Plant inoculation.
Fresh conidia were harvested from OTA medium with sterile water containing 0.02% Tween-20 as described 51 . For virulence detection, about ten-day-old rice seedlings were spray inoculated with conidial suspension 51 . Four pots (about 20 seedlings in a pot) were treated with each strain. For microscopic observation, the hollow space of rice leaf sheath from 5-week-old plants was inoculated with conidial suspension 52,53 . In order to examine the expression of MoNLP genes during infection of rice, the detached leaves of about 4-week-old rice were placed in plates with moist filter paper, and 20 μl conidial suspension (1 × 10 6 conidia ml −1 or 5 × 10 4 conidia ml −1 ) was dropped on the leaves. All pots and plates were placed in a moist plastic box at 25 °C for the first 24 hours in darkness, and then transferred to a chamber with a photoperiod of 16 hours under fluorescent light.
For fungal biomass assay, genomic DNA of infected leaves was isolated, and the amount of fungal DNA was quantified with qPCR as previously described 54 .
Data Availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.