Rice is one of the most studied plants for several reasons, one of which is its economic significance. Major pathogens to rice include Magnaporthe grisea, the causative agent of rice blast and Xanthomonas oryzae pv. oryzae, the causative agent of bacterial blight. Because these diseases inflict rice, extensive research has been done to understand host-pathogen interactions and develop disease management strategies. As a result, most rice resistance genes have been cloned against these diseases (Results Section). In this study, we embarked to understand BPB, a rice disease caused by another pathogen, B. glumae. To obtain a broad representation of the interaction between rice and B. glumae, we used RNA-Seq to identify transcripts that were differentially expressed between a resistant and a susceptible interactions. With young stems from seedlings as test tissues, we observed symptom development after 24, 48, 72 and 96 hours of inoculation. Stems over panicles were chosen for better distinction of symptoms between a susceptible and a resistant interaction over a short period of time. Reports have shown that symptoms on the panicles usually appear after two weeks of inoculation [3, 4]. At this time period, gene expression may not be indicative of what happens early in the interaction, but earlier time points may not show significant difference in the symptoms in the panicles. In the stem tissues, browning and lesion formation progressed at a remarkable rate in the susceptible genotype, especially when compared to the control and the resistant interaction. However, we found that after 48 hours was the earliest time point that showed the most difference in the responses between the two genotypes. This was our basis for choosing 48 hours post inoculation for the RNA-Seq experiment. We opted for replicates over deeper sequence coverage to provide statistical measures on comparative analysis between any two sample points as we were ultimately interested in possible sources of resistance genes from rice. Haas et al.
 expressed the same sentiment when they said that for some systems, the tradeoff for having replicates rather than sequence depth will provide better biological insight and statistical confidence. We had three biological replications and each replicate was loaded into two lanes. We pooled the reads from two lanes to constitute the read counts for a specific replicate (multiplexed with other samples). The application of the appropriate bioinformatic tools was equally necessary. In this work, we used Bowtie and rpkmforgenes.py script using default parameters to pre-analyze our data, and DESeq to calculate for the differentially expressed transcripts. High-throughput data analysis requires accurate prediction of variability within the dynamic range of values and a suitable error model and DESeq attempts to achieve them by using the negative binomial distribution with the variance and mean linked by local regression. DESeq was preferred over other programs because it provides better statistics for high-throughput data with few numbers of replicates such as RNA-Seq and it addresses the issue of data normalization in a more robust way compared to other available systems . As an initial filter after alignment into the rice genome, we eliminated reads that mapped to more than one locus, rRNAs and other repetitive sequences such as transposable elements. DESeq calculates padj (p-values adjusted for false discovery rate) to correct for multiple testing. We applied an FDR cutoff of 5% and used fold change values of 2 or greater to select for differentially expressed transcripts in the host (rice). These stringent conditions nevertheless generated sufficient transcripts for analysis.
The rice genes that were differentially expressed between the resistant and susceptible interaction were disease resistant-types or related, different enzymes, transcription factors, expressed and hypothetical proteins as well as proteins of unknown function. An enumeration of the transcript-types did not show significant distinction between the two types of interactions. However, a GO analysis, which classifies genes into biological process, molecular function and cellular component, demonstrated a clear distinction between them. In the R interaction, we saw an up-regulation of transcripts that are enriched for defense response, programmed cell death and generalized cell death transcripts under biological process; ATP, nucleoside, nucleotide and protein binding transcripts under molecular function; and mitochondrion supporting transcripts for cell component. By contrast, the S interaction displayed an up-regulation of transcripts enriched in lipid metabolic process, defense response and programmed cell death under biological processes; signal transducer, receptor activities and carbohydrate binding under molecular functions; and plasma membrane and membrane parts under cell component. This side by side comparison of ontologies presented that although disease resistance transcripts and most likely proteins were also up-regulated in the S interaction, other constituents appeared to play important roles in the resistance mechanism. For example, transcripts supporting molecular function and cell component were different between the two interaction types. Of note, a QTL for bacterial seedling rot, another rice disease caused by B. glumae is qRBS1, was mapped in chromosome 10 . This locus was not differentially expressed in any of the comparisons made, suggesting different pathways for resistance in these two rice diseases even if they were caused by the same pathogen. When we looked closer into specific disease-related transcripts, none of the previously cloned rice genes were differentially expressed, suggesting the known resistance genes were not involved in this interaction. Most were not even expressed at the tested conditions. The differentially expressed transcripts may represent genes that are unique to the rice - B. glumae interaction, indicating that resistance of rice to BPB may be conferred by a different set of genes and their roles in the interaction need to be further investigated. Of characterized resistance-related transcripts, those of the NBS-LRR-type and in some cases, sub-families such as NB-ARC and RPM1, were found to be both up-regulated and down-regulated in the R vs S comparison (down-regulation means up-regulation in the S vs R comparison), though actual transcripts associated with each group were not shared. We selected a few of these transcripts to verify their expression using qRT-PCR. All six disease-related transcripts (four RPM1 and two NBS-LRR types) were co-expressed in both the control and inoculated R conditions but not expressed at all in the control or inoculated S. A PIF-like ORF1 that is mapped in chromosome 8 follows the same trend. These results suggest that the R genotype maybe keeping a constitutive level of resistance arsenal to help it combat future B. glumae attacks.
It has been documented that the nucleotide-binding site or NBS (also NB-ARC) is a conserved domain for ATP binding and hydrolysis and sequences at the amino terminus are required for protein-protein interaction [48, 49]. The leucine rich repeats (LRRs) vary in number and the amino terminal domain seems to regulate activation while the carboxy terminal domain appears to function in recognition . It implies then that where NBS-LRR resistance genes are involved, so does specificity in the interaction. More so, it also suggests the involvement of an effector protein, which initiates the cascade of events that will lead to resistance. RPM1, a type of NBS-LRR resistance gene was originally cloned from Arabidopsis in response to the bacterium Pseudomonas syringae
. Prior studies had shown that the NBS-LRR gene family is constitutively negatively regulated [63–66] and gets activated in the presence of pathogens through a mechanism that is not clearly understood. However, their activation needs to be precise (in space and time) for resistance to ensue. In addition, earlier studies have demonstrated indirectly that the NBS motif binds to ATP or GTP for activation [48, 49, 64]. If this gene family functions in the same manner in this pathosystem, then processes necessary for their activation should be up-regulated or activated as well. Our GO annotation results suggest an enrichment of ATP binding activities under molecular function, supporting the premise that an activation of NBS type motifs occurs in the R genotype during B. glumae challenge. No evidence for NBS activation was shown in the inoculated S genotype, despite the up-regulation of this type motif.
The rice genome has been annotated with more than 500 NBS-LRR-type genes although more than a hundred were predicted to be pseudogenes [46, 47]. Available literature shows that they cluster where mapped  and high sequence diversity exists in both the NBS and the LRR domains . It has been proposed that this gene family arose by several independent events of gene duplications all throughout rice evolution . Prior research also demonstrated that diversifying selection has shaped the evolution of the family, giving rise to the diversity that has been observed among its members . Hence, it is conceivable that of those members that are functional, the mechanisms that they provide may not necessarily be similar. Our results showed a clustering of up-regulated transcripts in chromosomes 11 and 8 in the R vs S comparison. Although previous work showed a bias clustering of disease-related genes in chromosomes 11 and 12 , our results suggested that resistance against BPB was not a direct result of clustering alone as none of the resistance genes in chromosome 12 were differentially expressed between the conditions tested. Furthermore, there are other loci where resistance genes are clustered, though to a lesser degree . Our results suggest that rice may have utilized the clusters of resistance genes together with another factor/s to devise a resistance mechanism against BPB. These factors may include NBS-LRR activation partners like ATP binding. Another interesting finding was the co-expression in the water- and pathogen-inoculated states, and probably, constitutive expression of a PIF-like ORF1 transcript in the R and almost none in the S genotype. Because the filtering method for the reads that we used involved elimination of those mapped to more than one locus and known repeats like rRNA and transposable elements, we can state that the reads only mapped to the PIF-like ORF1 transcript in the genotypes that we studied. When we further investigated it from the Rice Genome Annotation Project , it showed only one match (located in chromosome 8), suggesting that only one copy exists per haploid genome. P instability factor or PIF
[52–56], a family of Class 2 transposable element is widely distributed in plants and other metazoans [55, 57, 58]. Jiao and Deng  performed a genome-wide survey of transcriptional activity of transposable element-related genes in 15 developmental stages and stress conditions in rice and found no expression of PIF-like transcripts in their test plants, suggesting that the PIF-like ORF1 is not expressed in all rice genotypes. PIF has two open reading frames, ORF1 and ORF2, of which ORF2 is most likely the transposase TPase . The function of ORF1 is still unknown, but its predicted protein sequence has significant homology to the Myb/SANT domain. The Myb domain is involved in DNA binding  while the SANT domain, although shares a strong homology with Myb sequences, is involved in protein-protein interactions . When we searched for homologs of the transposase in our transcriptomes, we found out that the reads mapped to several loci and were eliminated from the analysis. Whether the homologs were truly repetitive or this was an artifact of high-throughput sequencing analysis remains to be explored and is beyond the scope of this study. Based on the GO result that the R genotype did not show significant (p ≤ 0.05) enrichment for signal transduction, it appears that the PIF-like ORF1 may have been recruited to behave as a transcriptional regulator through DNA binding and not as a participant in signal transduction processes in this pathosystem. The transcripts that were constitutively expressed in the R genotype as quantified using qRT-PCR were all mapped in chromosome 11, suggesting that the PIF-like ORF1 (chromosome 8) may be acting in trans on the genes that it regulates.
All things considered, we propose a resistance mechanism in rice against BPB that existed early in rice domestication and that is not shared with other diseases including rice blast and bacterial blight. This was supported by the recent occurrence of this disease  and several observations that we noted in this work that are linked to resistance. We propose that shortly before it is domesticated, encounters between rice and B. glumae are limited. The genome of rice along with the prevailing environment at that time may have supported resistance. Specifically, the cluster of resistance genes that include NBS-LRR and related types in chromosomes 11 and 8, the up-regulation of the PIF-like ORF1 and the enrichment for ATP binding all contribute to this resistance. Because they are available, rice may have co-opted them as resistance contributors against BPB. The involvement of NBS-LRR-type transcripts and activation partner ATP binding suggests that the resistance mechanism consists of an effector molecule, probably from the pathogen, that is recognized by the host. The effector activates a cascade of events that will eventually lead to resistance in the host. It is possible that the PIF-like ORF1 may have been recruited to participate in the activation of the NBS-LRR genes. However, changes in global weather patterns, specifically gradual warming, favored the breaking of the resistance originally held by wild rice species. This is not outrageous as an increase in new or previously insignificant plant diseases caused by pathogens that grow optimally at higher temperatures has been observed with the increase in global temperatures [1, 71]. B. glumae is one of these pathogens.
Alternatively, because it may have been a part of a DNA transposon, the PIF-like ORF1 may be performing a more active role in the resistance pathway. This remains to be tested but is not a part of this study. Furthermore, we do not exclude that other processes may be occurring in parallel. The list of differentially expressed transcripts includes proteins of unknown functions and other disease related proteins. Their roles in the resistance pathway need to be uncovered in order to paint a complete picture of the resistance mechanism.