NIS is rare, weak, and isolate-dependent in temperate japonica rice genotypes
Partial resistance/susceptibility to rice blast was estimated under two nitrogen fertilization condition levels (low-level: N0; high-level: N1- see Methods) from a least-square means (Lsmeans) analysis of the density of blast lesions. We found 117 temperate japonicas (Additional file3) genotypes that were susceptible to the blast fungus isolate CL26, but no evidence of a global effect nitrogen fertiliser treatment on susceptibility (Additional file4). There was a significant genotype by nitrogen treatment interaction, with four genotypes (3%) showing an increase of susceptibility under nitrogen treatment and ten (8%) showing a decrease.
We found 159 rice lines susceptible to the blast fungal isolate CD203. In this analysis, the nitrogen fertiliser treatment, genotype, and the genotype-by-treatment interaction were all significant (Additional file5). The significant nitrogen treatment by genotype interaction implies that the impact of nitrogen on rice blast susceptibility depends on the rice genotype in this panel and thus that these data are suitable for a genome-wide association analysis.
The average increase in susceptibility to CD203 under the N1 condition was 8% (Fig. 1). Nineteen genotypes (11%) were significantly more susceptible in N1 compared to N0 conditions and five were significantly less (3%). The maximum increase in lesion number was observed in the TRAMONTO genotype, with four times more lesions in N1 condition than in N0. Conversely, the MARENY genotype showed three times fewer lesions in N1 condition than in N0 condition. Thus, our results indicate that NIS is relatively rare (14% of the panel) and weak (8% average increase) in this temperate japonica population. Heritability of Lsmeans for rice blast resistance was 0.43 and 0.42 for N0 and N1 conditions respectively (Additional file 6), allowing for GWA analysis within each fertilization condition.
LSmeans of N1 and N0 condition were used for the calculation of a new NIS Index (NISI, see Methods). This index allows the detection of genotypes for which the change of susceptibility upon fertilization deviates from the norm of 8% found at the level of the panel (Fig. 1). A genotype with a NISI of 1 has an increase of susceptibility equal to the global increase (8%). More robust genotypes will have NISI scores approaching zero, while less robust genotypes will have scores greater one. Twenty-seven percent of the genotypes showed a rather robust susceptibility with a NISI between − 1 and 1, and thirty percent had a NISI higher than three (Additional file7).
GWA analysis of rice blast resistance under different nitrogen fertilization
The GWA analysis identified fourteen significant SNPs in the N0 environment and three SNPs in the N1 environment (Fig. 2A and 2B respectively, and Additional file 8). To define a QTL, we only considered loci detected by several closely linked significant and sub-significant SNPs (see Methods). This excluded five SNPs found on chromosome seven that appeared to be scattered, with no other sub-significant SNP nearby. By contrast, the single SNP found on chromosome 5 appeared to represent a locus (NIS2) containing four sub-significant SNPs. Another locus, RRobN1 on chromosome 6, was defined by eleven significant SNPs.
Twenty-one significant SNPs were identified in the genome-wide association using the NIS Index score as the phenotype (heritability of NISI = 0.1) (Fig. 2C and Additional file 8). Among them, the four SNPs mapping on chromosome 2 and three on chromosome 3 were not considered further as they were scattered all along the chromosomes. A QTL (NIS3), defined by a block of 12 significant SNPs, was identified on chromosome 10. It is noteworthy that this QTL was not detected in the GWA analysis for lesion density, further demonstrating the usefulness of the NIS Index.
RRobN1, a QTL conferring partial resistance not affected by nitrogen fertilization
The RRobN1 (Resistance Robust to Nitrogen 1) locus on chromosome 6 was initially identified in the GWA analyses by eleven SNPs significant in the N0 but not the N1 condition, although many of the SNPs at this locus were sub-significant under this condition (Tables Additional file 8 and 9; and Fig. 3A). The haplotype associated with resistance is found in only four varieties of the panel: IRAT 268, IAC 26, GIGANTE VERCELLI, and RUBI. However, an ad hoc analysis considering only this locus suggests that it also conferred resistance under in the N1 environment (Fig. 3A). The level of resistance conferred by this locus was similar to other known major resistance genes , with a reduction in lesion number of 75% in both N0 and N1 conditions. However, RRobN1 does not confer resistance to the CL26 isolate (Additional file 10). Thus RRobN1 is isolate-specific and confers robust resistance under high nitrogen, two characteristics found for classical major resistance genes , suggesting that this locus underlies a classical resistance gene. The RRoN1 locus (650 kb between 23.90 to 24.2 Mb) on chromosome 6 does not contain any mapped or cloned resistance genes  and among the forty-two genes in the region there are no resistance gene analogs in the reference genome Nipponbare (Table Additional file 11). Thus the robust, elevated resistance associated with the RRobN1 locus may not be conferred by a classical resistance gene analog. Two phospholipidase D genes involved in disease resistance [56, 57] could be good candidates, among others.
NIS2, a QTL associated with increased susceptibility under high nitrogen fertilization
A second locus on chromosome 5, named NIS2, was only detected in the N1 condition against the CD203 isolate. There was no significant association of NIS2 with the CL26 isolate (Additional file 10). NIS2 was defined by a unique SNP in a strongly linked LD block (LD r² around 0.9 from 27.55 to 27.96 Mb) containing five sub-significant SNPs (with –log10 P greater than 4) (Fig. 3B). In fact, possibly two haplotypes could be defined at this locus, with the first haplotype (NIS2-1) found in 22 varieties (26% of the population). Plants with this NIS2-1 haplotype showed a 17% increase in lesion number from low to high nitrogen conditions (Fig. 3B). The second haplotype, NIS2-2, showed a more modest increase of 7%, similar to the average effect of nitrogen in the population (Fig. 1).
In order to investigate the underlying functional basis of the phenotype conferred by the NIS2 locus, all 69 genes present at this locus were investigated (Additional file 11). Among the many possible candidates, three functional hypothesis were retained. First, the increase of susceptibility could be due to the presence of regulators of biotic stress responses like RACK1B , OsNINJA1 [59, 60] and OsSYP71 . Second, the phenotype could arise from antagonistic pleiotropy in a gene such as OsMATE2 [61, 62], that regulates plant growth in response to nitrogen while negatively affecting disease resistance. Third, the phenotype could be due to genes modulating metabolism such as OsMYB55, which modulates amino acid metabolism , or OsNADH-GOGAT2 which is involved in nitrogen metabolism .
NIS3, a QTL conferring partial resistance strongly impacted by nitrogen fertilization
In contrast to RRoN1 and NIS2, NIS3 was identified using the NIS index (NISI; see Methods and above). NIS3 is located in an LD block of 800 Kb on chromosome 10 (Fig. 3C and Tables Additional file 8 and 9). We could define three haplotypes at this locus using the nineteen SNPs available in the 3000 rice genomes (Table Additional file 13). We could not further characterize the NIS3-2 haplotype because it was represented by only four genotypes in our panel. The NIS3-3 haplotype was the most represented in our sample (112 genotypes or 79%) whereas the NIS3-1 haplotype was less frequent, present in 25 genotypes. The mean NISI value of lines carrying the NIS3-1 haplotype was 3.46, which was significantly greater than lines with the NIS3-3 haplotype, which had a mean NISI value close to zero (Fig. 3C). NIS3-1 genotypes were more resistant than NIS3-3 genotypes, but also more sensitive under high nitrogen, with an increase of susceptibility of 14.6% (Fig. 3C).
We built a new rice panel (See Methods) in order to specifically test if haplotypes at the NIS3 loci were predictive of the phenotype across rice diversity. We chose twenty-six genotypes from different subspecies: sixteen with the NIS3-1 haplotype and ten with the NIS3-3 haplotype (Table in Additional file 2). This panel was inoculated with the CD203 isolate under the same conditions as those used for GWA. This experiment showed a significant interaction between nitrogen and genotypes (p-value = 4.33e-4) but no global nitrogen impact (p-value = 0.56). We did not use the NI because there was no global nitrogen effect in the ANOVA (NI ~ 0, thus NISI has to be adjusted; see Methods). As we observed in the original GWA, plants with the NIS3-3 haplotype were more robust to the effect of nitrogen treatment, with a mean adjusted NISI score of 0.68, compared to the adjusted NISI score of 3.46 for plants with the NIS3-1 haplotype. These results validate the involvement of NIS3 in NIS to the CD203 isolate (Fig. 4). There was no significant difference between the adjusted NIS index of haplotypes NIS3-3 and NIS3-1 (Additional file 14) in the panel of 117 genotypes susceptible to CL26 strains.
In order to understand the possible genetic basis of the observed increased susceptibility in the high nitrogen environment, the 126 genes annotated at NIS3 were investigated (Additional file 15). We considered as good candidates genes either involved in biotic stress response (twenty-nine pathogenesis-related genes), abiotic stress response (OsTIP3;1,  and RIPER6 ), the regulation of both stresses (OsGRXS17,  and OsTPR1 ), or nitrogen metabolism (OsPORB,  and a polylpolyglutamate synthetase ). The possible modulation of the M. oryzae pathogenicity program in response to a high nitrogen environment was also a hypothesis based on our previous results . Interestingly, OsTPR1 codes for a protein that competitively binds the fungal effector MoChia1, allowing the free accumulation of chitin and the re-establishment of the immune response during rice M. oryzae interaction . Finally, the NIS response could be due to a buffering of the immune response by primary metabolism at this locus. In that respect, OsPORB and the polylpolyglutamate synthetase genes have been shown to be repressed during blast infection .
The NIS3-3 allele is common in temperate japonica (79% of the panel tested, Additional file 16). We searched for interesting allelic variants of genes at the NIS3 locus that could have been fixed in this population. We found that an allele of RIPER6 (Ripening-related family protein), which is associated with cold tolerance , is observed in 90% of the cases in association with NIS3-3.