The Xoo strains used for genome sequencing in this study were collected over several decades from different rice planting areas in China ( Table S1 ). For comparison, the represented Xoo strains belonging to 4 different races from Japan and 6 different races from Philippines were also used. The genomic DNA of each Xoo strain was extracted and purified from the overnight using the Easy-DNA kit (Invitrogen, USA), following the manufacturer’s protocol. And then the genomic DNA were sequenced by using Illumina HiSeq 2000, 2500, or 4000 to generate paired-end reads with lengths of 100, 125, or 150 bp. For each Xoo strain, raw reads were assessed with the FastQC tool (https://github.com/s-andrews/FastQC) and quality filtered using Trimmomatic (https://github.com/timflutre/trimmomatic) and bases with low quality were discarded. The filtered reads were error-corrected by library with Quake to produce clean reads.

To study the population structure of Xoo from China, reads from each strain sequenced was mapped onto the reference genome PXO99^A, which is used as a model strain for studying the interaction between Xoo and rice host and is the first Xoo with complete genome sequence (Salzberg et al., 2008), using BWA (Li and Durbin, 2009). Variant detection was performed using Samtools mpileup and filtered with a minimum mapping quality of 30. SNPs were excluded if the coverage was less than 10% or more than 200% of the average coverage for the genome, or if not supported by at least 5 reads on each strand. Phage regions and repetitive sequences of the PXO99^A genome were predicted by PHASTER (http://phaster.ca) and RepeatScout (https://bix.ucsd.edu/repeatscout), respectively. Snippy (https://github.com/tseemann/snippy) was used to generate the whole genome alignment of all studied Xoo, and then the recombination was detected by Gubbins (https://sanger-pathogens.github.io/gubbins) on the core genome alignment after prophage and repeat regions were filtered. SNPs located within phage regions, repetitive sequences or recombinant regions were excluded.

To study the population structure of Asian Xoo, genomes from different locations were downloaded from the Genbank as follows: 100 from India, 9 from Philippines, 1 from South Korea and 1 from Japan (Quibod et al., 2016; Midha et al., 2017). We performed genome assembly from all the strains sequenced in this study. For each strain, the pair-end clean reads were assembled using Spades (Bankevich et al., 2012). After few evaluations, different sets of k-mer values were chosen for Spades according to different read lengths. The final assemblies were obtained by filtering out contigs with few reads supported or with lengths lower than 200 bp ( Table S1 ). Whole genome alignment was carried out on all the genomes obtained from public database and assembled here by Parsnp (Treangen et al., 2014) using the genome of PXO99^A as reference. Core SNPs were extracted from the core genome alignment by snp-sites (https://github.com/sanger-pathogens/snp-sites). Recombination of the core genome alignment was inferred by Gubbins. The final SNPs were obtained by filtering out the prophage region and repeat sequences mentioned above, and the recombined regions predicted here.

The SNPs-based maximum likelihood (ML) phylogenies were built by RAxML with the generalized time-reversible model and a Gamma distribution to model site-specific rate variation (Stamatakis, 2014). Bootstrap support values were calculated from 500 replicates. Xoo population structure of China was defined with the hierBAPS module of the BAPS software, which delineates the population structure by nested clustering (Cheng et al., 2013). Three independent iterations with upper population sizes of 5, 10, and 15 were used to obtain optimal clustering of the population. Phylogenetic trees were annotated and visualized by iTOL (https://itol.embl.de/).

To assess the virulence of Xoo, we used six near-isogenic rice lines, each containing a specific R gene, namely IRBB2 (Xa2), IRBB3 (Xa3), IRBB4 (Xa4), IRBB5 (xa5), IRBB14 (Xa14), and IR24 (Xa18), with IR24 serving as a susceptible check. All lines were obtained from the China National Rice Research Institute (CNRRI). After three weeks of sowing, seedlings were uprooted and transplanted into the field, with standard rice plant management techniques being applied. Xoo strains were cultivated on NA plates at 28°C for 48 h and suspended in sterile ddH₂O with concentration adjusted to 3×10⁸ cfu/ml before inoculation. For testing virulence, we used the leaf-clipping method, inoculating 15 top fully expanded leaves per rice plant with each strain. The length of the lesions was measured 21 days post-inoculation, and the ratio of the lesion length compared to the whole leaf length was calculated as previously described. Rice lines with a ratio lower than 1/4 were classified as resistant (R), and those with ratios between 1/4 and 1 were rated as susceptible (S). Furthermore, we measured the lesion length 14 days after inoculation and compared these values with those obtained seven days later, leveraging the method outlined by Liu et al. (2007).

Sequence data generated for this study have been submitted to the NCBI Sequence Read Archive under BioProject accession PRJNA350904.

To understand the population structure of Xoo in China, genomes of the 247 strains were sequenced, including 237 isolated from China, 4 and 6 representative ones from Japan and Philippines, respectively ( Table S1 ). Reads were mapped to the genome of PXO99A, SNPs were called by Samtools. After filtering out prophage region, repeat sequences and recombinant regions, we finally obtained a core SNP set containing 5,271 variable sites. Two strategies were used to outline the population structure based on the core SNPs, Maximum likelihood (ML) phylogenetic analysis and a tree-independent hierarchical Bayesian clustering (BAPS).

Based on phylogenetci analysis and BAPS results, all Xoo from China were clustered into 6 lineages, which were designated CX-1 to CX-6. Among them, more than 70% strains belonging to Lineages CX-5 and CX-6 ( Figure 1 ). The three major rice production areas, including South China, Yangtze Valley area and North China, were displayed in Figure 2A . We annotated the lineage-specific tree with the time and space information of the isolated Xoo ( Figures 1 , 2B, C ). All Xoo strains but one belonging CX-1 and CX-2 were isolated from infected rice leaves from Yunnan province of South China in the year of 2003. The Xoo belonging to CX-3 were isolated from North China in 1984, 2003 and 2014, with most of which from Northeast China in 2014. Xoo belonging to CX-5 and CX-6 were widely dispersed in China during the past 30 years; while most members of CX-5 were dominant in South China and Yangtze Valley, and CX-6 were more frequently isolated from North China and Yangtze Valley. It is suggested that although bacterial blight diseases in rice from different regions of China were caused by different Xoo lineages, CX-5 and CX-6 showed a greater ability to spread among different regions, indicating stronger adaptability than the other lineages.

To investigate more details about evolution and dynamics of Xoo in China, we focused on the two major lineages, CX-5 and CX-6. Phylogenetic analysis based on core SNP alignment ( Figures 3A, B ) and pairwise SNP distance between isolates was performed for both lineages ( Figure 3C ). Though no correlation between root-to-tip branch lengths and the known years of isolation of the sequenced Xoo was observed for both lineages, some important points about the evolutionary dynamics of these two lineages were illuminated, especially those for the recent sporadic outbreaks leading to bacterial blight diseases on rices at a short time and in a small-scale area.

Four sub-lineages were identified in CX-5, among of which, CX-5.1, CX-5.2, and CX-5.5 were restricted in South China, while CX-5.3 and CX-5.4 can be found in some places of both South China and Yangtze Valley ( Figure 3A ). When time was considered, we found that all these sub-lineages persisted at these places at least for decades. Then we focused on the recent sporadic outbreaks being attributed to this lineage. Besides clustered together on the tree, we considered epidemiologically and genomically linked outbreak with an average SNP pairwise distance less than 5. In 2003, the outbreak in Hunan province of Yangtze Valley was caused by CX-5.4 (HN-2003), while the outbreak in Guangdong province of South China ascribed to two sub-population of CX-5.5 (GD-2003-1 and GD-2003-2). One cluster of CX-5.3 caused outbreak in Guangxi province of South China in 2014 (GX-2014).

CX-6 was divided into 5 sub-lineages, CX-6.1 to CX-6.5 ( Figure 3B ). Seven out of 10 isolates in CX-6.1 were from different places of South China. Actually, this sub-lineage was the most diverse one in CX-6 with long branches. In CX-6.2, Xoo from North China and Yangtze Valley were dominant on different branches, respectively. CX-6.3 and CX-6.4 were most frequently found in Yangtze Valley, while Xoo from North China predominated in CX-6.5. Similar as CX-5, all these sub-lineages have contributed to BB on rice for a long time in China. These sub-lineages totally caused at less 4 local outbreaks in 2003 and 2014. In 2003, the local outbreaks in two provinces of Yangtze Valley (Henan, and Jiangsu) were caused by CX-6.2 (HeN-2003) and CX-6.4 (JS-2003), respectively, while the one in North China (Jilin province) was ascribed to CX-6.5 (JL-2003). Outbreak in Jiangsu province (Yangtze Valley) in 2014 was caused by one sub-population of CX-6.4 (JS-2014).

There are two major domesticated rice subspecies Oryza sativa japonica and indica existing in different areas of China; and most of rice varieties were derived from these two subspecies (Kovach et al., 2007; Huang et al., 2012). Rice planted in high-altitude areas of Yunnan province from South China and North China belong to subspecies japonica (http://www.ricedata.cn). Xoo isolated from these areas were respectively clustered into 3 lineages, CX-1, CX-2, CX-3 and one sub-lineage of CX-6 (CX-6.5) ( Figures 1 , 3B ). Lineage CX-5 mainly contained Xoo isolated from rice in South China where subspecies indica is planted ( Figure 3A ). CX-6 also contained Xoo isolated from rice of japonica from Yangtze Valley (CX-6.4), indica from Yangtze Valley (CX-6.3) and some indica from South China (CX-6.1), indicating the widest adaptation to different rice varieties of Xoo in this lineage ( Figure 3B ). In CX-6.2, old Xoo were dominant on indica, while recent isolated ones were almost from japonica, suggesting a putative host switch in this sub-lineage.

Taken together, the distribution of Chinese Xoo lineages appeared to be impaired by both biogeography and rapid dynamics over isolated time, which may be mainly due to the distribution and dynamics of the two major subspecies of rice, japonica and indica.

To place Xoo from China into a global context, phylogenomic analysis was performed on the core genome SNPs of 109 Xoo genomes available in Genbank (including 100 from India, 8 from Philippines and 1 from Japan), and the 247 sequenced ones in this study. Each of the draft genome assemblies of these Xoo strains has a genome size of around 4.5 Mbp similar as reported previously which is about 500 Kbp less than that of the complete genome, mainly causing by huge repeat elements in the Xoo genome. The GC contents of these genome sequences are between 63% and 64%. Phylogenetic analysis revealed a similar topological structure of Asian Xoo to that of Chinese Xoo, with some small differences in some branches ( Figure S1 ). Eleven genetic lineages were identified for the Xoo population from Asia. Xoo from India and Philippines were respectively assigned into PX-A to PX-C and IX-I to IX-V according to previous studies (Quibod et al., 2016; Midha et al., 2017). Eight lineages displayed strict geographic distribution features, most of them being specific to one or at least a limited amount of rice-planting areas. Besides the three Chinese lineages, CX-1, CX-2 and CX-3, PX-B and PX-C were mainly represented in Philippines, and IX-I to IX-III were from India. CX-4 contained old Xoo from China and some recently isolated ones from India belonging to IX-V. For CX-5 and CX-6, besides both represented the major Xoo from China, they also contained a few Xoo from Japan, India and Philippines, with PX-A of Philippines being nested within the former one and IX-IV from India being encompassed among the members of the last one, indicating frequently transmission of these 2 lineages not only among different places of China but also among different Asian countries.

To determine the pathogenic diversity of Xoo strains, virulence assays were conducted on six rice lines. Based on the interactions between Xoo strains and rice lines, most of the tested Xoo were classified into 9 pathogenic races, designated as R1 to R9 in this study ( Table S1 ; Figure 4A ). The top two races with most members, R5 and R8, contain 26% and 18% of the isolates, respectively.

The three major rice-planting regions had different race compositions ( Figure 4B ). Yangtze Valley mainly contained Xoo from races R5 and R8, with the respective percentage of the total population being 50% and 30%, respectively. Xoo isolates from North China mostly belongs to R7 and R5, together accounts for about two third of the total population. Five of the nine Xoo races were well represented in the Xoo population from South China, with each of them account for more than 10% of the total population. The virulence dynamics of Xoo during the last 30 years was hypothesized by analyzing race distribution in China at different period ( Figure 4C ). Strains isolated before the year of 2000 were mainly of the races of R5, R4, R8 and R1, while most Xoo isolated in 2003–2004 were R8, R7, R5 and R6; and about 45% of the isolates in 2014 were R5.

We also found different lineages have different race compositions, indicating genetic background playing important roles in the interaction between Xoo and rice hosts ( Figure 4D ). Compared to other lineages, the CX-6 Xoo isolates fall mostly within the races of R5, R8 and R4, while have reduced representation in R1, R2 and R6. Moreover, the rapid dynamics of pathogenic features can be inferred by focusing on the races of Xoo isolated from the same places at the same time and belonging to the same lineage ( Figure 1 ). For example, most Xoo from Yunnan, one province of South China, in 2003 were clustered in CX-1 and CX-2. The 19 isolates in CX-1 represented 8 of the 9 races described above, and the 14 strains in CX-2 belonged to 6 races. Even Xoo isolates from the same outbreak were allocated into different races, with 16 Xoo in JL-2003 belonging to 3 different races and 10 isolates in GD-2003-2 representing 2 races.

As each of the near-isogenic rice lines used in this study contains one known resistance (R) gene, we studied the interaction between Xoo and rice host by focusing on the capacity of Xoo to overcome the resistances conferred by these genes ( Figure 1 ; Tables 1 , 2 ). When the time of isolation was considered, we found that Xoo from different times have different capacity to overcome different R genes ( Table 1 ). R gene xa5 showed the most effective resistance against Xoo all the time. At different periods, more than 90% Xoo could not infect the rice line IRBB5, which contained gene xa5. The resistance mediated by R gene Xa3 was effective during the period of 1990 to 2004. However, rice containing this R gene only showed resistance against 5% of Xoo in 2014. Rice cultivars carrying R gene Xa4 are resistant to more than 60% of Xoo isolated from 1970s to 2000 and 2014. However, they are only resistant to about 30% of the isolates recovered in 2003 and 2004.

To determine whether the Xoo genetic background contributes to the interaction between Xoo and rice lines, we studied the resistance capacity of R genes against Xoo from different lineages ( Table 2 ). We found that resistance provided by R genes against Xoo are indeed depended on the genetic background of Xoo. R genes Xa2 and Xa14 effectively confer resistance against Xoo from Lineages CX-1, CX-2 and CX-3, preventing infection by more than 70% of Xoo strains from these lineages. Similarly, R gene Xa3 and Xa4 showed potential resistance against Xoo from CX-1 and CX-2, and CX-2 and CX-3, respectively. No R gene except xa5 conferred resistance to more than 50% of Xoo strains from CX-5 and CX-6, which is consistent with the situation that CX-5 and CX-6 had much more extensive distribution across different areas of China during the past decades, comparing to the other Chinese Xoo lineages. Among the sub-lineages of CX-5 and CX-6, Xa2 was only effective against Xoo of CX-6.5, with effectiveness of 90%, while Xa4 could confer the resistance against more than 60% Xoo from these two lineages except for CX-5.1 and CX-6.5; Xa14 showed potential medium level of resistance against Xoo from CX-6.1 and CX-6.5 ( Table S2 ).