Plants were sown on Wednesdays in Neuhaus S soil in which poudzolane was added (2L/70L). Standard fertilization solution containing nitrogen source (75% NO3-/25% NH4+; 40 mg/L) was supplied every Monday for 3 weeks. Twenty six days after sowing, we supplied on Monday either a fertilization solution containing a nitrogen source (1N condition: 50% NH4⁺/50% NO3⁻; 40 mg/L), or the same solution without nitrogen source and corresponding to the 0N condition. This fourth fertilization was done 1 day before inoculation.

Plants were inoculated with spore suspensions as described in Berruyer et al. (2003). For RNAseq experiments, we used GUY11 isolate as it is a reference for many M. oryzae experimental studies. For gene expression studies, we also used a mock treatment corresponding to the solution into which spores are re-suspended (i.e., 0.5% gelatin solution). Five to seven days after inoculation, the symptoms were scanned and the number of susceptible lesions were counted using ImageJ software (http://rsbweb.nih.gov/ij/) as described in Ballini et al. (2013).

Plant tissues were harvested before and after fertilization. They were oven dried about 48 h at 65°C. Total nitrogen was analyzed in the samples by the Cirad analysis laboratory US49 in Montpellier (France). The content in nitrogen was established by a thermoelectric cell after burning in a furnace at 900°C under an oxygen flow according to the Dumas method. Amino acids and sugar contents were quantified as described in Gravot et al. (2010).

RNA extraction was performed as mentioned in Delteil et al. (2012). Quantitative PCR was performed using LC 480 SYBR Green I Master Mix (Roche, Basel, Switzerland) and a LightCycler 480 instrument (Roche). The amount of plant RNA in each sample was normalized using actin (Os03g50890) as internal control. PCR primers are provided in Supplementary Table 1.

Tissue staining was performed as described previously (Ballini et al., 2013). Observation were performed on LSM700 (ZEISS, http://www.zeiss.com) confocal microscope.

Total RNA was isolated with TRIzol (Invitrogen) from each sample according to manufacturer's instructions. It was treated with RNase-free DNase according to manufacturers's instructions (Promega) to remove residual DNA.

For RNA-seq, total RNA samples were sent to BGI Tech (Shenzhen, China) for sequencing. The libraries were sequenced as 101-bp paired-end reads using Illumina Hiseq2500 according to the manufacturer's instructions. Clean reads have been mapped using SOAP2 on reference genomes: Kasalath for rice and 70–15 for Magnaporthe. Annotated rice transcripts from Kasalath genome were downloaded from the TASUKE web-platform (Kumagai et al., 2013; Sakai et al., 2014). M. oryzae annotation datasets were downloaded from the Magnaporthe comparative Sequencing Project at the Broad Institute (assembly release 8; Chiapello et al., 2015).

The RNA sequencing depth allowed a good coverage of rice and M. oryzae genes. For one repetition, on average 15.9 × 10⁶ reads had a perfect match on Kasalath genome (non-inoculated or mock inoculated samples) and 46.1 × 10⁶ reads (inoculated samples). Up to 8,002 of the 31,507 Kasalath rice genes identified by the reads mapping had a low coverage (<5 reads) and were removed from the analysis in order to increase the statistical power. The sequencing depth allowed the detection of a large number of M. oryzae genes: on average 14,208 reads perfectly matched in an annotated gene in 0N samples and 37,864 reads in 1N samples. This difference is due to the higher amount of symptoms and thus of fungal RNA in the 1N condition. However, 7,324 genes had a low coverage and were removed in order to increase statistical power. FastQ files, reads counts and RPKM per gene are available on GEO database (GSE83219).

Differential expression between all repetitions was performed on DEB website (Yao and Yu, 2011) for each of the four conditions 0N mock inoculated, 0N Guy11 inoculated, 1N mock inoculated and 1N Guy11 inoculated. All the data were normalized together. For differentially expressed gene analysis we used DESeq with a minimum FDR of 5%. We considered as differentially expressed genes (DEGs) all genes that showed a significant difference in at least one of the four conditions. In order to conduct enrichment analysis we transferred when possible annotations from Nipponbare to Kasalath. We also used the Archipelago database referencing defense-related genes defense regulators (Vergne et al., 2008). Nitrogen pathway was annotated manually based on function annotation. Gene expressed after glutamine treatment were annotated based on GEO microarray GSE56770 (Kan et al., 2015). ABA responsive genes were collected based on literature (Gómez-Porras et al., 2007). Enrichment analysis in the different functions was tested against whole genome annotation with a Chi square test corrected by Bonferroni. GSEA was performed against whole genome on the rice array database (http://www.ricearray.org/).

For Differentially expressed gene analysis we used two different statistical analysis methods, namely DESeq and EdgeR with a minimum FDR of 5%. We considered as differentially expressed genes (DEGs) all genes that were significant in at least one of the statistical test. In order to perform enrichment analysis we did some complementary annotations. Gene coding for Carbohydrate-Active enZYmes (CAZY) were annotated based on CAZY.org database (Lombard et al., 2014). Pathogenicity genes were annotated based on literature. Small secreted proteins were annotated based on three criteria: (1) the presence of a signal peptide longer than 15 amino acids at the beginning of the protein, (2) the absence of a transmembrane (TM) domain or a TM domain in the first 10 aa and (3) a protein size smaller than 250 aa. Signal peptide was detected using SignalP program (Petersen et al., 2011). Transmembrane domains was detected using TMHMM program (Krogh et al., 2001). Enrichment analyses was performed on Fungidb database (Stajich et al., 2012) and for the selected functions the frequency was tested against the whole genome annotation with a Chi square test corrected by Bonferroni.

We have previously described a laboratory protocol inducing NIS in the Kasalath but not in the Nipponbare variety after inoculation with M. oryzae (Ballini et al., 2013). Plants were grown during 4 weeks and 1 day before inoculation they are separated in two sets: fertilized without nitrogen (0N: all macro and micro-elements except N) or full fertilization (1N). An increase of leaf elongation was visible at 2 dpi in Kasalath (NIS) but not Nipponbare (no NIS) (Supplementary Figure 1). This increase was moderate and almost undetectable at the end of the experiment, when disease symptoms were developed (5 dpi). Thus, our experimental system minimizes the impact on plant growth.

Several evidences suggest a correlation between total nitrogen content and blast susceptibility (Otani, 1959; Talukder et al., 2005). Therefore, we measured total nitrogen content in rice leaves before and after infection in plants grown under 0N and 1N regimes (Supplementary Figures 2A,B). Total nitrogen content was not affected in Nipponbare before or after inoculation. However, in Kasalath total nitrogen content was significantly higher by up to 8.5% under the high nitrogen regime at 2 dpi. Thus, a slight increase in total nitrogen content is observed during the fungal invasive phase in NIS. To investigate in more detail the changes in nutrient contents, total amino acids and three sugars were quantified in rice leaves in three independent experiments (Supplementary Figures 2C–F). Before inoculation, no significant change could be observed between 0N and 1N samples in Kasalath or Nipponbare (Supplementary Figures 2C,D). In Nipponbare no significant change could be observed at 2 dpi although there was a tendency toward less metabolites detected in 1N samples. On the contrary, at 2 dpi a significant difference was observed in Kasalath for glutamine and alanine in mock inoculated plants but not in inoculated plants. The overall tendency in Kasalath for the other metabolites was an increase in 1N samples. However, further analysis will be necessary to establish if the changes in glutamine contents in M. oyzae inoculated samples could be causal for NIS.

In our experimental system, disease symptoms on Kasalath plants, as measured by the number of susceptible lesions, were significantly increased with high levels of nitrogen fertilization (Figure 1, Supplementary Figure 3A). We reproducibly observed from 2 to 6 times more lesions for the Guy11 isolate and also an increase of lesion size, up to 2.2-fold (Supplementary Figure 3B). This is consistent with what was found by Otani (1959). In contrast, disease severity in the Nipponbare genotype against the Guy11 isolate was not affected (Figure 1, Supplementary Figures 3, 4).

Our NIS system had not been tested in other rice/isolate combinations. To evaluate if the observed effects of nitrogen supply were specific to the Guy11 isolate, we further tested five virulent rice blast isolates (Figure 1, Supplementary Figure 4). Nitrogen supply did not trigger any significant increase of susceptibility in Nipponbare. By contrast, disease severity in Kasalath was higher with the six isolates of M. oryzae tested after nitrogen supply (Figure 1, Supplementary Figure 4). The increase in lesion number triggered by nitrogen supply varied from a maximum of 17-fold with the VT2 isolate and a minimum of 5-fold with the CL32 isolate. This protocol also induced enhanced susceptibility in a large range of rice genotypes (Ballini et al., 2013). Thus, our experimental NIS protocol has a minimal impact on plant development and allows studying the mechanisms of NIS in a panel of rice variety/M. oryzae isolates with differential response to nitrogen.

Up to now, there has been no evaluation of the effects of nitrogen supply on the growth of the blast fungus before and after penetration in the rice leaf. We monitored the growth of the isolate Guy11 in Kasalath and Nipponbare plants at different time points after inoculation (1, 2, and 4 dpi) and under the 0N and 1N conditions. As shown in Figure 2A, we observed no difference in Nipponbare between the 0N and 1N conditions. This is consistent with the absence of increased susceptibility in this cultivar (Figure 1). In contrast, several significant differences were found in Kasalath plants (Figure 2B). First at 1 dpi, 85% of the spores developed its specialized penetration cell, appressorium, under 1N conditions compared to 70% under 0N conditions. Thus, the fungus seems to grow faster on the leaf surface of plants that were supplied with nitrogen. This difference in fungal growth is no longer observed at 2 dpi as the fungus has penetrated in 60–65% of the cases, irrelevantly of the nitrogen regime. Thus, the final penetration efficiency does not seem to be affected by nitrogen supply. Finally, at 4 dpi, before the first symptoms appear on the leaf surface, we could measure a significant difference in the percentage of cells that each individual colony had invaded. Whereas, 45% of the fungal colonies invaded several cells under the 0N condition, more than 80% could be observed under the 1N condition. Therefore, the growth of the blast fungus inside the plant cells is higher after nitrogen supply. Based on these results we decided to further examine the molecular changes occurring at 2 dpi, when fungal growth does not seem to be significantly affected and before the strong changes observed later on at 4 dpi.

In order to investigate the transcriptional regulation during NIS in Kasalath inoculated with Guy11, we decided to conduct a dual RNA-seq experiment. In this way we were therefore able to follow both rice and M. oryzae gene expression before (only for rice) and 2 days after inoculation. The data are available on GEO database (GSE83219).

At 0 dpi RNA-seq analysis on rice genes allowed the detection of 243 differentially expressed genes (DEG): 65 are more expressed in 1N and, 178 are less expressed in 1N (Supplementary Table 2). The 65 highly expressed genes are enriched in defense-responsive genes present in the rice defense database (Vergne et al., 2008). Thus, this analysis shows that there is no massive modification of rice transcription 1 day after nitrogen supply in our condition, although some defense regulators may be induced.

At 2 dpi, differentially expressed genes were detected by comparing four different possible situations: No nitrogen and mock inoculation (0N Mock), nitrogen-fertilized and mock inoculation (1N Mock), no nitrogen and Guy11 isolate inoculation (0N Inoc) and last nitrogen-fertilized and Guy11 isolate inoculation (1N Inoc; Figure 3, Supplementary Table 3). We considered a gene as differentially expressed (DEG) when DESeq analysis gave an adjusted P < 0.05. Previous studies indicate that the transcriptome of combined treatments is poorly predicted by the transcriptome of single treatments (Rasmussen et al., 2013; Suzuki et al., 2014) and we made a similar observation. For instance, a total of 2,711 DEGs were identified with DESeq analysis between “1N Inoc” condition and “0N Mock” condition while only 1,848 of these genes (68%) could be predicted when analyzing the four conditions separately.

Using our statistical analysis on the four conditions separately, we could identify a total of 2,665 DEGs in six groups (Figure 3). We compared all group of genes (list of genes in Supplementary Table 3) showing similar expression pattern with available known groups of genes involved in disease resistance (Vergne et al., 2008 and updated versions) or in nitrogen/glutamine response (GEO database; GSE56770) and enrichment values are given in Supplementary Table 4.

The first group represents 637 nitrogen-specific DEGs that are genes differentially expressed between 0N and 1N conditions independently of infection: 222 are induced and 415 are repressed by nitrogen (Figure 3A). No detectable enrichment in gene related to nitrogen pathway was observed in this experiment, which suggests that transcriptional changes following fertilization may have happened earlier than 2 dpi to increase the observed nitrogen content (Supplementary Figure 2). The second group represents 1,237 DEGs specific to infection by M. oryzae (Figure 3B) as they are differentially expressed during infection independently of the nitrogen treatment: 903 are induced and 334 are repressed by infection. As expected, this group of DEGs is enriched in defense-related genes like PR proteins. While these first two groups represent classical nitrogen-related and infection-related situations, the four remaining groups (representing 789 DEGs; Figures 3C–F) contain genes that are differentially expressed by both nitrogen and infection conditions.

The third group gathers 29 DEGs genes whose expression is normally modified by nitrogen or infection but is globally conserved between 0N and 1N during infection (Figure 3C). Infection may prioritize the expression of these genes despite a negative cross talk with nitrogen regulation. The final level of expression of these genes is conserved at 2 dpi whatever the fertilization used. This group is enriched in WRKY transcription factors and response to biotic stress.

The fourth group represents 707 DEGs differentially expressed during infection with an amplified response to inoculation in 1N condition (Figure 3D). This family is enriched in PR proteins, defense regulators, phytoalexin pathway, WRKY transcription factors and response to biotic stress. Several known positive and negative defense regulators are part of this group like some regulators of transcription (OsWRKY62, OsWRKY28, OsNAC4, OsbHLH148, OsSpl7, NH1, and OsNPR4), an allene oxide synthase involved in JA biosynthesis (OsAOS2), a salicylic acid 3-hydroxylase (OsS3H) and several kinases (OsBSR1, OsBIDK1, OsMPK5). Thus, the transcription of defense-related genes and regulators is amplified in 1N condition at 2 dpi although plants are finally more susceptible (see Discussion). Interestingly this group is also enriched in glutamine responsive genes like the OsGS1-2 gene coding for a glutamine synthetase (Supplementary Figure 5; Kan et al., 2015). This is in accordance with previous results showing that OsGS1-2 is expressed at a lower level in rice roots grown with low nitrogen (Funayama et al., 2013; Yang et al., 2015). This observation was followed-up by the analysis of a rice gs1-2 mutant during NIS (see below).

The fifth group (Figure 3E) is composed of 16 genes which response to inoculation is suppressed in 1N condition. The sixth group is composed of 37 genes (Figure 3F) that are repressed during infection under the 1N condition only. These two last groups potentially represent functions that are impacted negatively by high nitrogen and that may otherwise favor resistance; however, there was no massive enrichment of genes involved in disease resistance in this group. Instead we found several genes that are related to senescence like OsNAP and OsTZF1. The OsNAP transcription factor, whose induction by inoculation under 0N is abolished in 1N (Figure 3E), has a pivotal role in senescence regulation (Liang et al., 2014) and jasmonic acid biosynthesis (Zhou et al., 2013). Similarly, the OsTZF1 transcription factor that is repressed in infected plants under 1N but not 0N (Figure 3F) is known to inhibit senescence and to induce defense-related genes and HR-like lesions (Jan et al., 2013). Thus, the fifth and sixth DEG groups gather genes that could be implicated in NIS regulation based on their expression profiles. As expected, genes involved in nitrogen assimilation and remobilization are among these groups (Supplementary Table 3) and we followed-up the hypothesis that such genes are key determinant of NIS.

Glutamate metabolism is not only crucial for amino acid metabolism but it has been reported as a key player in defense (Seifi et al., 2013). In their review Seifi et al. have reported that glutamate metabolism functions in opposite ways: fueling plant defense or diverted by the pathogen to facilitate infection. They identified three major pathways that may lead to resistance or susceptibility: Nitrogen remobilization away from infected tissues, TCA cycle replenishment and cell death. We used this review to identify the genes associated with these three defense strategies that were identified in our RNA-seq data set using annotation (Figure 4) and looked if these pathways were affected in infected plants under high nitrogen. We identified 18 genes implicated in nitrogen assimilation in our dataset: 15 are not differentially expressed in none of the four considered conditions at 2dpi. OsGS1-2 is the only gene whose response to M. oryzae is amplified by nitrogen treatment (Figure 3D). This expression pattern is not correlated with the detected level of glutamine in the leaves at 2 dpi (Supplementary Figure 2F). When comparing plants grown under 0N and 1N regimes, an increase in glutamine level was observed in mock inoculated samples but not in M. oryzae inoculated samples. In the same experiment used for RNAseq analysis, OsGS1-2 expression increased at a higher lever under 1N regime than under 0N regime. OsNADH-GOGAT2 is nitrogen specific induced (Figure 3A) and a GDC (Os06g45670) is induced by M. oryzae at a level that is not significantly different in both N treatments. We identified 21 genes associated to several metabolism process that possibly link the cell death induction to GS/GOGAT pathway in our dataset: 17 are not differentially expressed in any of the conditions. OsNOA1 is repressed and OsPAO4 is induced during infection in both nitrogen regimes. OsOXO4 is the only gene whose transcriptional response to M. oryzae is amplified by high nitrogen. However, the role of OsOX4 in rice blast resistance is not clear (Zhang et al., 2013; Li et al., 2015). Finally, we identified in our datasets 11 genes coding for enzymes potentially involved in the TCA cycle replenishment or GABA shunt, 9 of which are not differentially expressed. OsGDH3 and OsGAD3 have an amplified response to M. oryzae in presence of nitrogen. OsGAD3 and OsGDH3 may be controlled by the pathogen in order to boost TCA cycle and compensate the drain of nutrient away of the infected tissues. In N and P deprived environments, OsGDH3 is induced in rice shoots probably to boost nitrogen recycling (Qiu et al., 2009). Thus, we did not detect a massive transcriptional reprogramming of the three pathways identified by Seifi et al. (2013) during infection under high nitrogen. However, the modification of the regulation of some of the key enzymes by infection could explain NIS and to test this hypothesis, we analyzed the role of OsGS1-2 in defense.

The GS enzyme catalyzes the conversion of glutamate to glutamine using NH4+. In rice there are three genes coding for isoenzymes of cytosolic glutamine synthetase (OsGS1-1, OsGS1-2, and OsGS1-3). One of these genes appears as a good candidate for NIS mechanisms investigation: OsGS1-2, which was specifically induced in infected plants under high nitrogen (Supplementary Figure 5). This gene, is important in the primary assimilation of NH4+ taken up by rice roots (Yamaya and Kusano, 2014). The knock-out gs1-2 mutants showed reductions in the content of glutamine, glutamate, asparagine, and aspartate, and an increase in free NH4+ ions in the roots, shoots and xylem sap (Funayama et al., 2013). We obtained a knock-out mutant in the OsGS1-2 gene using a line where a T-DNA is inserted at the end of the coding region (Supplementary Figure 6A). The expression level of OsGS1-2 was strongly reduced (Supplementary Figure 6B) while plant growth was not affected in this gs1-2 mutant. Under our growth conditions, we were able to detect a slight decrease in fructose, glucose and some amino acids in the shoots of the gs1-2 mutant compared to the wild type (Supplementary Figure 7). Infection assays with M. oryzae of wild-type and gs1-2 mutants were done under low and high nitrogen. We observed that the gs1-2 mutants were more resistant than wild-type plants (Figure 5). More importantly, while wild-type plants showed a typical NIS phenotype, the gs1-2 mutants did not show an increased susceptibility under high nitrogen fertilization.

The dual RNA-seq experiment allowed us to follow fungal gene expression in planta at a time (2 dpi) where we did not measure any difference in fungal growth inside tissues from rice fed with or without nitrogen (Figure 2). Of the 60 million reads sequenced in each of the six experiments at 2 dpi, ~26,000 perfectly matched M. oryzae genes and not rice. We decided to compare the expression between the two nitrogen treatments for genes with a minimum coverage of 5 reads (3428 genes out of M. oryzae 10,752 genes detected; Supplementary Table 5). Overall, this set of 3,428 genes expressed in planta at 2 dpi is significantly enriched in genes coding for SSP (small secreted protein: 293 genes; 8.5 against 6.2% expected from genomic data) and known pathogenicity genes (2.1 against 1.3%) but not in cell-wall degrading enzymes (CAZY) (4.2 against 3.8%) (Supplementary Table 6), consistent with previous results of M. oryzae expression in planta (Matsumura et al., 2003; Mosquera et al., 2009; Kim et al., 2010; Mathioni et al., 2011; Kawahara et al., 2012). Only a few genes (24) were significantly differentially expressed between our 0N and 1N conditions and most had unknown functions (Supplementary Table 7). In order to get further insight of the global trends of gene expression in M. oryzae, we analyzed the global and not the individual fold change of some functional categories amongst the 3,428 investigated genes. In this way, we could detect tendencies in the pattern of gene expression of specific biological or molecular functions. Of the 293 genes coding for SSPs, 131 were more expressed in 1N condition (Log₂ fold change >0,7) whereas only 32 were less expressed in 1N condition (Log₂ fold change < −0,7) (Supplementary Table 5). This tendency was not observed for other pathogenicity genes, vesicle trafficking components, transcription factors or CAZY enzymes (Supplementary Table 5). In order to validate this tendency, we tested the expression of a set of 19 genes coding for SSPs and other pathogenicity factors by Q-RT PCR in a biological replicate of the RNA-seq experiment (Kasalath plants showing NIS). Seven of these genes were indeed differentially expressed as expected from RNA-seq data (Figure 6): Pmk1 (Thines et al., 2000), Bas2 (Mosquera et al., 2009), AvrPi9 (Wu et al., 2015), and three putative effectors (MGG_08715, MGG_15774 and MGG_08428) were more expressed whereas VPS39 (Ramanujam et al., 2013) was less expressed in 1N conditions. Interestingly, all these genes were less expressed in 1N condition in Nipponbare (where NIS was not observed). The 12 other genes tested were not differentially expressed in either variety (Supplementary Figure 8). This data suggests that M. oryzae can sense the physiological status of the plant and potentially adapt its pathogenicity strategy by delivering a higher amount of effectors in high nitrogen conditions.

Rice blast disease is highly dependent on the developmental stage of the plant (Vergne et al., 2010) but several results presented here with our NIS protocol argue against the possibility that NIS is an indirect effect of nitrogen on plant development. In our system, there was only a limited modification of plant growth after infection (Supplementary Figure 1). Our way of applying nitrogen for triggering NIS did not strongly affect plant physiology, as shown by total nitrogen and metabolome measurements (Supplementary Figure 2). Some amino acids like glutamine and alanine significantly but slightly increased in plant supplied with nitrogen and a similar trend could be observed for other metabolites. Plant transcriptome was also only moderately affected by nitrogen fertilization, as shown by RNA-seq data before infection or in mock treated samples at 2 dpi (Supplementary Tables 2, 3). Indeed, while infection alone affected more than 1,237 genes at 2 dpi, nitrogen alone only affected 243 genes before infection (Supplementary Table 2) and 639 genes at 2 dpi (Figure 3). Thus, based on physiological, metabolomics and transcriptomic results, we conclude that our NIS protocol is only slightly affecting plant growth and plant physiological processes in the absence of infection whereas it strongly impacts on disease severity. Further investigation is required to confirm if the slight increase in some metabolites (e.g., glutamine) could be causal of NIS.

Although a link between amino acids metabolism and defense response has been demonstrated several times, a remaining question is how nitrogen fertilization affects defense. Nitrogen fertilization and subsequent modifications of amino acid content could play a role in fueling the defense response (Masclaux-Daubresse et al., 2006; Bolton, 2009; Nussbaumer et al., 2015; Park et al., 2015) or in regulating key defense regulators (Rojas et al., 2014). Similarly to previous results, we observed in our dataset that defense-related genes (e.g., PR genes and genes involved in the biosynthesis of the antimicrobial secondary metabolites) are more expressed during infection under high nitrogen fertilization (Figure 3). However, our results also show that there could be a complex regulation of defense in nitrogen fed plants. Indeed despite the amplification observed for PR genes and eight positive regulators of defense, several negative regulator of defense were also strongly induced by high nitrogen (OsWRKY62, OsWRKY28, OsMAPK5, OsNPR4, OsS3H). Thus, we cannot exclude the possibility that among the many types of defense regulators induced by high fertilization, some negative components may be critical for increasing susceptibility. However, if this was the case, one would expect end-executers like PR genes to be down-regulated, which is not the case. More subtle cases could also explain how high nitrogen may impact on susceptibility, for instance the case of the induction of OsNAP gene by infection found to be abolished by high nitrogen (Figure 3E). OsNAP negatively regulates the biosynthesis of abscisic acid (ABA; Liang et al., 2014) and positively regulates the biosynthesis of jasmonic acid (Zhou et al., 2013). Since ABA and jasmonic acid are respectively a negative (Jiang et al., 2010) and a positive (Shimizu et al., 2013) regulator of rice resistance to blast, the induction of OsNAP could also participate indirectly to defense. The absence of induction of OsNAP by infection under high nitrogen would prevent the proper setting of immunity. Thus, based on our results, we propose that the increased susceptibility in nitrogen-treated plant cannot be explained by a general down-regulation of the defense pathway.

The idea that nitrogen fertilization increases nutrient availability and thus pathogen growth is supported by our results. First, it seems that fungal growth may increase on the plant surface under high nitrogen (Figure 2) at a time where there was no obvious down-regulation of plant defense (Supplementary Table 2). In other systems (for review see Robinson, 1980), it has been observed that germination and growth of the fungus on the plant surface was increased by high nitrogen levels. This suggested that the pathogen can sense on the leaf surface some metabolic modifications of its host consecutive to nitrogen fertilization, potentially amino acid levels. Fungal growth was also increased in infected plant tissues at 4 dpi (Figure 2). Similar results were observed for colonies of Blumeria graminis f.sp. hordei grown on barley treated with high nitrogen fertilization (Newton and Guy, 1998). Second, the gs1-2 mutants, which showed a reduction in sugars and amino acids in leaf tissues before infection, were more resistant to M. oryzae (Supplementary Figure 7). This decrease in some sugars and amino acids content is in line with other reports suggesting complex effects of OsGS1-2 function (Bao et al., 2014). Consistent with our results, several Arabidopsis mutants affected in the regulation of nitrogen and amino acid metabolism have been shown to be impaired for pathogen resistance: nrt2-1 (Camañes et al., 2012), nrt2-6 (Dechorgnat et al., 2012), amt1-1 (Pastor et al., 2014), atl-31, and atl-6 (Maekawa et al., 2012). The Arabidopsis lht1 mutant (lys histidine transporter 1) shows strong resistance to several pathogens that could be explained by a reduced cellular content in proline, glutamine and alanine (Liu et al., 2010). This is consistent with our observation that glutamine and alanine levels are increased in the case of enhanced susceptiblity in Kasalath plants under high fertilization (Supplementary Figure 2). Additional measurements are needed to establish whether modifications of metabolite fluxes and/or quantities are responsible for NIS. Infection by M. oryzae is known to modify the plant TCA cycle, in particular GABA-shunt, probably to divert the host metabolism (Wu et al., 2006; Takahashi et al., 2008). Based on our RNA-seq results (Figure 4), some key components of the TCA cycle that are differentially regulated (OsGAD3 and OsGDH3) could also participate to the fueling of the pathogen during infection under high nitrogen. Altogether, these elements support the hypothesis that nitrogen fertilization fuels growth of M. oryzae.

One final hypothesis could be that nitrogen fertilization could indirectly regulate fungal pathogenicity functions. Despite a large amount of data on M. oryzae effector and secretome (Matsumura et al., 2003; Mosquera et al., 2009; Kim et al., 2010; Mathioni et al., 2011; Kawahara et al., 2012), little is known on the regulation of M. oryzae effector by the physiological status of the plant. Recently, we have demonstrated that in rice plants recovering from drought, the expression of M. oryzae effectors was massively down-regulated (Bidzinski et al., 2016). More generally the regulation of plant pathogen effectors is still largely underscored (McCotter et al., 2016). We identified a set of M. oryzae genes regulated by the level of nitrogen fertilization applied to the host plant. Here we were able to demonstrate that M. oryzae effectors are more induced in nitrogen-fed Kasalath plants than in nitrogen-deprived ones, and thus that pathogenicity functions are potentially increased. However, according to available knowledge on fungus biology in vitro, a nitrogen-rich environment should turn down many pathogenicity genes (Donofrio et al., 2006; Wilson et al., 2012; Fernandez and Wilson, 2013). This apparently contradictory set of data could be explained by the fact that all previous studies were done in vitro whereas our present study was conducted in planta. Thus, NIS in Kasalath plants could be the consequence of an up-regulation of pathogen effectors under this condition. This is similar to the observation that in Cladosporium fulvum, the expression of the Avr9 effector is induced by nitrogen (Thomma et al., 2006). Our result thus extends this observation by showing that a massive re-programming of effectors can be controlled by nitrogen fertilization in planta.

To conclude, our results clearly indicate that nitrogen fertilization increases susceptibility despite an increase in the expression of several defense genes. This enhanced defense may be directly or indirectly due to an increase in some metabolites (e.g., Glutamine). At the meantime, the fungal pathogenicity program is also visibly up-regulated by nitrogen fertilization, although it is yet difficult to identify which metabolite is responsible for this behavior. We propose that after nitrogen fertilization, despite an increase in defense, the host does not succeed in facing a concomitant increase in pathogenicity of the fungus, leading to enhanced susceptibility The logic applied in this work allowed us to identify GS1-2 gene as a potential key player in NIS. These results open the way to possible improvement of disease resistance under high fertilization.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.