Rice Xa21 primed genes and pathways that are critical for combating bacterial blight infection

Rice bacterial blight (BB) is a devastating rice disease. The Xa21 gene confers a broad and persistent resistance against BB. We introduced Xa21 into Oryza sativa L ssp indica (rice 9311), through multi-generation backcrossing, and generated a nearly isogenic, blight-resistant 9311/Xa21 rice. Using next-generation sequencing, we profiled the transcriptomes of both varieties before and within four days after infection of bacterium Xanthomonas oryzae pv. oryzae. The identified differentially expressed (DE) genes and signaling pathways revealed insights into the functions of Xa21. Surprisingly, before infection 1,889 genes on 135 of the 316 signaling pathways were DE between the 9311/Xa21 and 9311 plants. These Xa21-mediated basal pathways included mainly those related to the basic material and energy metabolisms and many related to phytohormones such as cytokinin, suggesting that Xa21 triggered redistribution of energy, phytohormones and resources among essential cellular activities before invasion. Counter-intuitively, after infection, the DE genes between the two plants were only one third of that before the infection; other than a few stress-related pathways, the affected pathways after infection constituted a small subset of the Xa21-mediated basal pathways. These results suggested that Xa21 primed critically important genes and signaling pathways, enhancing its resistance against bacterial infection.

In light of the broad disease resistance spectrum and endurance as well as distinct metabolite profiles 24 , it is necessary to conduct an independent study that focuses primarily on the mechanisms of Xa21-conferred BB resistance. In this study, we exploited two isogenic rice genotypes, one with and the other without the Xa21 gene, profiled their transcriptomes before and within the first 96 hours after Xoo infection, and contrasted the variations of the transcriptomes of the two rice lines in reference to the transcriptome of the normal rice plants. The analyses of the large quantities of gene expression profiling data from deep sequencing and a de novo genome sequencing result revealed the genes, biological processes and signaling pathways that are responsible for the resistance to rice blight infection.

Results And Discussion
A system for study of Xa21 mediated BB resistance. To establish a platform for studying Xa21, we generated a nearly isogenic line (NIL) of rice indica variety 9311 that carries the desirable Xa21 gene. In order to minimize or eliminate the possible impact of diverse genetic background on gene expression, we used the 9311 rice, which is susceptible to Xoo, as the recurrent parent and the CBB23 rice, which carries Xa21, as the donor, and successively applied more than 15 generations of backcrossing to introduce Xa21 into 9311, creating a new line of 9311/Xa21 rice. Fifteen generations of backcrossing are more than double the minimum of 6 backcrossing generations typically required to recover the phenotypes of the recurrent parental line and eliminate most of the donor chromosome fragments linked with the target Xa21 gene. As a result, the 9311/Xa21 rice, referred to as the R(esistant) plant, should have nearly the same genetic background as the 9311 rice, denoted as the S(usceptible) plant, except that the former carried Xa21 and its linked genomic fragment.
The nearly isogenic property between the R and S plants was further tested and validated using multiple means, including phenotyping, genotyping, whole genome sequencing and transcriptome profiling. The inclusion of Xa21 in the R plants was first indicated by the substantially smaller lesions on the R plants than on the S plants after Xoo inoculation (Fig. 1A) and the presence of a molecular marker co-segregated with Xa21 in the R plants (Fig. 1B). Furthermore, three lines of evidence showed the similar genetic backgrounds of the R and S plants except a small region encompassing the Xa21 gene. The R and S plants shared many similar agronomic traits (Supplemental Table S1), exhibited similar AFLP (Amplified Fragment Length Polymorphism) profiles (Fig. 1C), and had nearly identical genome sequences, based on re-sequencing of their genomes (see Methods), on all chromosomes except a ~2 Mbp region encompassing Xa21 on chromosome 11 of the R plants (Fig. 1D). In short, these results showed that the R and S rice lines were nearly isogenic.
The ~2 Mbp long Xa21-hosting region, which was inevitably introduced into the R plants along with Xa21 through backcrossing, had a high rate of genotypic variation (Fig. 1D) and harbored a total of 250 genes, among which 90 had non-synonymous mutations. It is noteworthy that only one of the 90 non-synonymously mutated genes, except Xa21, was expressed. We suspect the functional impact of the genes in the Xa21-hosting region except Xa21 is minimal. In support of this belief, we showed in a previous comparable study that a transgenic rice line and a rice line generated by backcrossing, both of which carry Xa21, were substantially equivalent at the transcriptome level 25 . Moreover, as validated by real time PCR, Xa21 in the R plants was constitutively expressed across all of the eight time points before and after Xoo infection that we profiled (Fig. 1E). In contrast, the regulatory role of expressed Xa21 on downstream genes was implicated by transcription factor OsWRKY62, which had a much lower abundance in the R plants than in the S plants before Xoo infection (Table S2). OsWRKY62 directly interacts with the Xa21 protein and acts as a negative regulator against Xoo in rice by suppressing defense-related genes 11 . Together, these results clearly showed that most of the genetic backgrounds of the R plant had been substituted with those of the S plants by backcrossing, and the R and S plants formed an ideal system for studying the functions of Xa21.
Xa21 suppressed Xoo growth. Our first step to characterizing the functions of Xa21 was to profile the in planta growth of pathogen Xoo within both R and S plants in the first 10 days after pathogen inoculation (see Methods). No apparent difference of pathogen growth was observed between the R and S lines within the first 3 days while the pathogen was making its way into the host cells. However, starting on day 4 the growth of Xoo was substantially suppressed in the R plants with respect to the growth in the S plants (Fig. 1F). The suppression was statistically significant (t-test, p-value < 0.05) and persisted at least for 10 days after inoculation. This in planta observation revealed Xa21 as a suppressor of Xoo growth so as to curtail its virulence. Xoo triggered broad perturbation in rice transcriptomes. In order to investigate how Xa21 responded to Xoo infection, we adopted high-throughput deep-sequencing to profile the transcriptomes of both R and S lines before Xoo infection and at seven time points within the first 4 days after pathogen inoculation (see Methods). Sequencing profiling produced more than 594 million raw reads from 19 plant samples (Table S3). About 82.24% (± 1.29%) of the raw reads can be mapped to the rice reference genome and more than 92.10% (± 1.26%) of the qualified reads (after removing low-mapping-quality reads) can be mapped to the exon regions of annotated genes (Table S3). This sequencing-based profiling provided a deep and broad map of transcriptome variations of the R and S rice plants in the process of Xoo infection.
Gene expression profiles of biological duplicates of the R and S plants before Xoo inoculation were produced to assess the quality of deep-sequencing based gene expression profiling. The result showed that the expression profiles were highly reproducible, with the Pearson correlation coefficients between the duplicates of the R plants and between the duplicates of the S plants being 0.8532 and 0.8052, respectively ( Fig. 2A).
We analyzed gene expression levels to characterize transcriptome variations in the period of the first four days of infection (see Methods) between the R plants inoculated with Xoo and the 9311 rice (the S Remarkably, the two transcriptomic perturbations were incompatible, even if the difference between the two rice lines was only due to the ~2 Mbp Xa21-hosting region in the R plants. A total of 3,496 genes exhibited significant expression variations between the R and S plants for at least one of the seven time points after Xoo infection (Fig. 2C,D and Table 1), and most of these DE genes were aggregated into 187 of the 316 (59.18%) annotated pathways in rice (RiceCyc pathway database, version 3.3, Dharmawardhana,  Ren et al. 2013) (Table S4); these 187 DE-gene containing pathways were referred to as the Xa21-induced pathways. These results of distinct transcriptomic responses suggested that Xoo activated different signaling pathways in the two rice plants, resulting in their distinct Xoo resistances (Fig. 1A,F).
Since the R rice exhibits stable resistance to Xoo throughout all stages of rice development, it is viable to hypothesize that genes continuously up-regulated or down-regulated across the R and S plants were most likely to be related to the resistance of Xa21. Along the seven time points profiled during the infection of Xoo, 12 genes were consistently differentially expressed between the R and S plants, where 4 genes were highly induced and the remaining 8 suppressed in the R plants (Fig. 3A). Four of these 12 genes (LOC_Os02g18140, LOC_Os11g36180, LOC_Os11g35710, LOC_Os11g36160) were further examined by real-time PCR for their expression variations across the R and S plants before Xoo inoculation (Fig. 3B). Note that LOC_Os02g18140 encodes a NBS type disease resistance protein and the other three genes are on chromosome 11 where Xa21 resides. LOC_Os11g36160 and LOC_Os11g36180 were particularly interesting since they reside in the neighborhood of Xa21, were highly induced in the R plants, and were annotated to be receptor kinases.
Xa21 mediated complex basal signaling pathways to prepare for Xoo infection. Besides the broad, distinct perturbations to the transcriptomes of the R and S rice caused by Xoo infection, the most surprising result of the transcriptome profiling was a great deal of transcriptomic difference between the R and S rice before bacterial infection. Precisely, 1,889 genes, involving in 135 signaling pathways, exhibited significant expression variations between the R and S plants before Xoo inoculation (Fig. 2E). This is remarkable as it clearly indicated that Xa21 was already functional before the infection. These 135 pathways, referred to as the Xa21-mediated basal pathways, were related to various types of material and energy matabolisms (Table S4). Among them, 28, 26 and 4 Xa21-mediated basal pathways were related to basic material and energy metabolisms, cellular components, and synthesis metabolisms, respectively. In contrast, based on Gene Ontology, many Xa21-induced processes after infection were directly related to stress responses and infection (Table 2), including all kinds of phytohormones and phytoalexins, whereas the first stress related biological process was only ranked the 24th among the Xa21-mediated basal processes (Table S5).
It is worthwhile to mention that at any of the seven time points after Xoo infection that we profiled, the degrees of differential expression between the R and S plants were substantially less than before the infection (Fig. 2E, the first columns of all time points). For example, at 8 hours post inoculation (hpi), 116 genes were differentially expressed, which were only 6.25% of the 1,889 DE genes before the infection; these 116 DE genes were involved in 10 signaling pathways (Table 3), in sharp contrast to the 135 Xa21-mediated basal pathways. On the other hand, approximately 66.6% of the Xa21-mediated basal pathways overlapped with the Xa21-induced pathways (Table S4). This high degree of overlap suggested that, even before Xoo infection, Xa21 had prepared the R rice well so that it was able to respond to the presence of Xoo as effectively and quickly as possible, as illustrated by its strong resistance to Xoo (Fig. 1). Together, these results suggested that before Xoo invasion, Xa21 in the R rice effectively reallocated energy and resources among many house-keeping cellular activities to prepare the plant, and after infection, Xa21 adjusted or activated stress related pathways according to a given Xoo strain and the time of pathogenesis.
Cytokinins contribute to Xa21 resistance to Xoo. The most notable enriched pathways between the R and S plants at various time points after Xoo inoculation were the ones related to phytohormones (Table 4 and S4). The pathways related to jasmonic acid (JA) and ethylene (ET), two classic phytohormones, had many DE genes in two of the eight time points profilied. Meantime, the pathways of some other members of hormone families, such as brassinosteroids (BR), gibberellic acid (GA) and cytokinins (CK), were also enriched ( Table 4). Some phytohormones regulated themselves through different pathways at different time points. For example, at 4 hpi, the GA biosynthesis pathway had a substantial number of DE genes, however, at 24 hpi, the pathway for gibberellin inactivation was enriched. Furthermore, different DE genes were involved in the same pathways of the same hormones and showed distinct expression patterns in the R and S plants at different time points in the first four days of infection (Table  S4).
Among the phytohormones detected by the transcriptome profiling, CK was the most pronounced. The DE genes residing on the CK pathways appeared at five of the seven time points after Xoo infection, four of which were the highest ranked among all pathways detected (Table S4), alluding to CK's involvement in Xa21-mediated resistance to Xoo. These CK pathways included that for cytokinins-O-glucoside biosynthesis (PWY-2902), cytokinins-9-N-glucoside biosynthesis (PWY-2901), and cytokinins-7-N-glucoside biosynthesis (PWY-2881), which were responsible for CK conjugation and thus for inactivating cytokinins 26 . CK is known to be involved in plant disease resistance 27 and in biotic 28 and abiotic 29-31 stress responses. The DE genes on the cytokinins-glucoside biosynthesis pathways exhibited a relatively coherent pattern of expression (Fig. 4A). Before Xoo infection, most of the DE genes, i.e., 15 of the 16 (93.75%), had lower abundances in the R plants than in the S plants. In contrast, at the 12, 24 and 72 hpi, 26 of the 27 (96.30%) DE genes on the cytokinins-glucoside biosynthesis pathways had higher abundances in the R plants than in the S plants. At 96 hpi, all (100%) of the 18 DE genes on the cytokinins-glucoside biosynthesis pathway had lower abundance in the R plants. For example, the low expression abundances of 4 genes in the CK-related pathways were validated by real time PCR (Fig. 4B). Such a high degree of coherent expression pattern over the course of Xoo infection suggested that the role of endogenous CK in Xa21-mediated resistance is complex and is difficult to be imitated by a constant amount of exogenous CK or by constantly inducing/repressing CK-related genes, which was often adopted in the previous studies on the functions of endogenous phytohormones.

Pterocarpan phytoalexins might function to repress Xoo in the R plants. Maackiain, together
with medicarpin, is the main pterocarpan phytoalexin in chickpea and occurs exclusively in a (-)-(6aR,11a R)-configuration 32 . Two adjacent genes on rice chromosome 1, LOC_Os01g01650 and LOC_Os01g01660, encode the enzymes for production of medicarpin and maackiain, respectively. These two genes had similar expression patterns in the R and S plants before Xoo infection. At the initial Xoo infection, their expressions were dramatically repressed at 4 hpi in both R and S plants. The suppression of these two genes persisted in the S plants throughout the 96 hours of infection. These results suggested that Xoo suppressed the synthesis of pterocarpan phytoalexins to promote its growth and disease progression in the S plants. However, the expressions of these two genes in the R plants elevated dramatically after 12-hpi and peaked at 24-hpi, which were further validated by real time PCR (Fig. 5A,B). The expression levels were suppressed again after 24-hpi and dropped to nearly 0 at 96-hpi. The increased expression of the two genes in the R plants might promote the synthesis of pterocarpan phytoalexin to control BB disease within the early stage of 24-hpi.
Notably, these two genes were not highly expressed in a rice line that is susceptible to rice blast fungal infection after the fungal infection (data not shown), suggesting that pterocarpan phytoalexins responded differently to fungal and bacterial infections. In addition, mock infection with water did not induce these two genes in the S plants either, confirming that they responded to Xoo infection.

Involvement of iron in Xa21-mediated disease responses. Iron is a key nutrient for bacterial
growth, and the usable form of iron for microorganisms is usually siderophore 33 . As Xoo colonizes within rice xylem, where siderophores are derived, the host typically exploits the essentiality and toxicity of transition metals to defend against bacterial invaders 34 . Before Xoo infection, the pathway of the enterobactin biosynthesis, a catecholate siderophore, was significantly enriched with 12 DE genes (FDR = 0.08703, Table S4), among which, 11 genes had lower abundance in the R plants. The experimental validation of three of the DE genes, using real-time quantitative PCR was consistent with the result from sequence profiling (Fig. 5C). The reduced expressions of these genes on the enterobactin biosynthesis pathway might help lower the amount of siderophores, which in turn restricted Xoo colonization in the R plants after Xoo infection. Because of iron uptake deficiency, a mutation in the Xoo feoB gene causes severe virulence deficiency and constitutive production of a siderophore 35 . A defect in siderophores formation in Dickeya dadantii, a plant soft-rotting enterobacterium, leads to symptoms localized to inoculated leaves, indicating that the siderophores are required for bacteria to spread to the other parts of the plant 36,37 . The R plants had limited blight lesions on inoculated leaves (Fig. 1), indicating that Xoo was curtailed in the R plants. Since most known mechanisms of disease resistance through iron-withholding are realized by regulation of iron-binding proteins 33,38,39 , it is viable to hypothesize that siderophores were also restricted in the R plants through iron-withholding. The way to restrict active iron in the form of siderophore seemed to be rare; this may be a double-edged sword because the lack of active iron is harmful to Xoo as well as to the rice plant at the same time.
Concluding remarks. The 9311 rice (the S line) and its nearly isogenic line with the Xa21 gene (the R line) that we used form a robust tool for studying the function of Xa21. In combination of deep sequencing-based transcriptome profiling and bioinformatics analysis, our results provided remarkable genome-wide profiles of gene expression and related signaling pathways and biological processes that significantly differed in the two rice genotypes. The significant difference between the transcriptomes of the two rice genotypes before Xoo infection revealed insights into the functions of Xa21 in priming various metabolic pathways so as to gain high and durable resistance to Xoo. After Xoo infection, Xa21 mediated DE genes and pathways were sharply reduced but more related to resistance to Xoo. Among them, the plant hormones, especially cytokinins, were broadly involved, suggesting complex mechanisms of hormones in Xa21-mediated resistance to Xoo.

Material and Methods
Rice varieties and growth condition. Rice 9311 variety, a popular indica rice restorer line 40    Pathogen inoculation and evaluation. Xanthomonas oryzae pv. oryzae (Xoo) Philippine race 6 (P6) was used for pathogen inoculation. Xoo was subcultured at 28 °C on PSA (Potato-Sugar-Agar) medium (potato, 300 g/L; Ca(NO 3 ) 2 •4H 2 O, 0.5 g/L; Na 2 HPO 4 •12H 2 O, 2.0 g/L; sugar, 15 g/L; agar, 15 g/L) for 3 days. Inoculums were prepared by suspending the bacterial cells in sterile water and adjusting the concentration to about 10 9 cells per milliliter. The last rice leaves were infected with P6 by using scissors dipped in bacterial suspensions to clip leaves 1-2 cm down from the tip of the leaf blade at the heading stage of 9311 and 9311/Xa21 41 . Mock-infected plants were treated in a similar fashion except that water substituted for P6. Fifteen days post inoculation, lesion length was measured from the cut surface at the tip to the distal-most position on the leaf that exhibited a grey, chlorotic or water-soaked lesion.
DNA and RNA isolation and genetic analysis. Fifteen randomly selected leaves were harvested at each time point (Table 1) and pooled to represent each treatment. After harvest, leaves were immediately frozen and stored in liquid nitrogen until use. About 100 mg samples were grinded to powder with liquid nitrogen for DNA and total RNA isolation using the Total DNA/RNA Isolation Kit (R6731, Omega, USA) following the manufacture's protocols. The total RNA quality was measured using Agilent RNA To investigate whether Xa21 was introduced into the genome of the R plant, we designed a molecular marker (Sequence listed in Table S6), named as U1/I2, which was co-segregated with Xa21. This marker was able to amplify a fragment of 575 bp in the R plant and a fragment of 445 bp in the S plant.
The method for analyzing the expression of Xa21 was described in our early report 42 ; Amplified fragment length polymorphism (AFLP) was used for genetic background analysis for the R and S genotypes, using the methods of Vos et al. 43 . AFLP primers were given in our earlier report 44 . Assay for quantification of bacterial growth. The bacterium population was determined using three P6-infected leaf samples collected at 0d (day), 1d, 2d, 3d, 4d, 6d, 8d and 10d after inoculation. The bacterial growth was analyzed according to Song et al. 45 with three biological replicates.
DNA and RNA library preparation, emulsion PCR and sequencing. About 1 μ g genomic DNA from the R and S plants were used for DNA library preparation, using the SOLiD ™ 5500 Fragment Library Core Kit (Part no. 4464412) according to its user guide. A total of 20 μ g total RNA was used for two rounds of mRNA purification using Dynabeads (610.06, Invitrogen, USA). About 100 ng mRNA was fragmented using NEB Next Magnesium RNA Fragmentation Module (E6150, NEB), purified with an RNA clean up kit (R6247, Omega, USA), end repaired with T4 Polynucleotide Kinase (T4 PNK) (M0201, NEB) and cleaned up again with a kit (R6247, Omega). The end-repaired RNAs were used to prepare the strand specific transcriptome, using the Small RNA Sample Preparation kit (E6160, NEB) according to the manufacturer protocol with some minor modification, including the SR Primer F3 being replaced with barcode primers. The resulting DNA and RNA libraries were used for emulsion PCR to produce the beads for sequencing on the SOLiD 5500 machine, using 75 nt mode and 75 nt + 35 nt mode for the sequencing of DNA and RNA libraries, respectively. Biological duplicates of RNA libraries of the R and S plants before Xoo infections were profiled for quality assessment.
Analysis of genotypic variation. The DNA sequencing reads in the color-space format were mapped to the Oryza sativa Nipponbare reference genome and gene annotation from MSU Rice Genome Annotation Project (Release 7 46 ) using LifeScope (Life Technologies) software version 2.5.1. Genotypic variations in the R rice line were analyzed using the genomic regions where the genome sequences of the S line and the 9311 reference genome were the same in order to rule out possible impact of natural mutations in the S plants. The analysis had a resolution of 5 Kbp, in which sequence variations within a 5 Kbp window were tallied. The calling of a genotypic variation at a genomic locus of the R line was subjected to a set of stringent criteria: the locus had distinct nucleotides in the R and S lines; and the sequencing must have at least 5X coverage of the same read at the locus to rule out or minimize possible sequencing error. Gene expression and differential expression analysis. For RNA analysis, the reads mapped to each annotated gene were tallied using the whole transcriptome analysis workflow of LifeScope. Differential expression analysis was performed using the edgeR 47 package. The reads count per gene was normalized using the TMM method in edgeR. An exact test, analogous to the Fisher's exact test, was performed based on the normalized counts with the common dispersion factor being set to 0.1. Genes were considered to be differentially expressed if they had more than 10 read counts across all samples and their False Discovery Rates (FDR) 48 of the exact test were no greater than 10%.  Table S6. Data were normalized using the reference gene LOC_Os06g11170.1 (coding for a putative nucleic acid binding protein) with the same primers published by Narsai et al. 51 . Before performing expression analysis, the primers' efficiency was estimated through a five-point standard curve. Only the target genes with amplification efficiencies between 90%-105% were chosen for expression validation. All PCR reactions were done in three biological replicas and three technical replicas. Ct values were exported and analyzed using Microsoft Excel 2010.