Genome-Wide Association Study Using Genotyping by Sequencing for Bacterial Leaf Blight Resistance Loci in Local Thai Indica Rice

Abstract

Bacterial leaf blight (BLB) is a devastating disease caused by Xanthomonas oryzae pv. oryzae (Xoo), which poses a significant threat to global rice production. In this study, a genome-wide association study (GWAS) was conducted using the genotyping-by-sequencing (GBS) approach to identify candidate single nucleotide polymorphisms (SNPs) associated with BLB resistance genes. The study utilized 200 indica rice accessions inoculated with seven distinct Xoo isolates and filtered highly significant SNPs using a minor allele frequency (MAF) of >5% and a call rate of 75%. Four statistical models were used to explore potential SNPs associated with BLB resistance, resulting in the identification of 32 significant SNPs on chromosomes 1–8 and 12 in the rice genome. Additionally, 179 genes were located within ±100 kb of the SNP region, of which 49 were selected as candidate genes based on their known functions in plant defense mechanisms. Several candidate genes were identified, including two genes in the same linkage disequilibrium (LD) decay as the well-known BLB resistance gene (Xa1). These findings represent a valuable resource for conducting further functional studies and developing novel breeding strategies to enhance the crop’s resistance to this disease.

1. Introduction

Rice (Oryza sativa L.) is one of the world’s most popular staple foods and is consumed by more than half of the world’s population. In 2021, almost 780 million tons of rice were produced, with about 90% of the production occurring in Asia [1]. However, current production levels are still insufficient to meet the demand for rice in Asia. Food security can be assured by preventing the impacts of climate change, such as pathogen spread. This is one of the fundamental problems in rice production resulting from biotic stresses, which were identified as a major problem [2,3,4]. World rice production is estimated to incur losses due to pests and pathogens, with bacterial leaf blight (BLB) being among the top five pathogens responsible for these losses [5].

BLB disease, caused by Xanthomonas oryzae pv. oryzae (Xoo), is one of the most serious diseases affecting rice production worldwide. In the past, there was confusion between two diseases: leaf streak and bacterial leaf blight. This was the case until 1957, when they were clearly classified. In 1978, Xanthomonas campestris pv. oryzae was identified as the causal agent of both diseases. However, in 1990, to alleviate confusion, the pathogen was renamed Xanthomonas oryzae pv. oryzae. [4,6]. The first report of BLB in rice was published in Japan in 1884 [5,7]. The disease then spread to other parts of Asia, including China, Korea, India, Indonesia, the Philippines, Taiwan, Thailand, and Vietnam, before spreading worldwide [6]. The first report of Xoo in Thailand was found in Bangkok, where it infected several Thai rice varieties, including Hom Set Ti, Khao Ta Hang, and Niew Kan Plu [8]. As a vascular disease, BLB can adversely affect rice yields. The losses could depend on the variety of rice, environmental conditions, and rice development stage [9,10,11]. The infection resulted in a systemic condition in which the plant leaves presented a tannish grey-to-white appearance with lesions occurring along their veins. The symptoms can be divided into three phases based on their characteristics, including the leaf blight phase, kresek phase, and pale-yellow phase, each with its own infection stage. Leaf blight is the most common symptom observed after plant transplanting and usually appears within the first 4–6 weeks following planting. The lesions appear as yellow or straw-colored stripes along the leaf veins, and the leaf edges are also affected, forming a curved margin from the tips to the leaf bases [11]. A small yellow droplet of bacterial ooze may be observed at the beginning of the blight phase. The second phase of seedling infection is known as the kresek phase, causing wilting of young plants during the early stages of infection. Once leaves become infected, they turn yellow or green and begin to roll up. Another phase involves pale-yellow leaves on the youngest parts of the plant, which become white or pale yellow in color [11]. As a result of the lesions on the leaf blade or leaf tips, BLB symptoms, such as yellowing, wilting, and drying, were easily identified.

The simplest method of preventing BLB is using chemical pesticides; however, widespread application of these chemicals poses a threat to rice food safety. Additionally, any chemical control implemented in a monsoon climate would be completely ineffective. In order to control this disease and ensure food security, resistant rice cultivars containing R genes must be developed and implemented [12,13,14]. According to reports in recent years, there have been more than 45 genes that confer resistance to BLB, which has increased over the years [15,16]. As far as R genes are concerned, they are mostly found on chromosome 11 [17,18,19,20]. However, R genes are also present on other rice chromosomes except for 9 and 10 [19,21,22,23,24]. Overall, several GWAS studies have been conducted to identify genes associated with BLB resistance in rice. Shu et al. [25] identified 58 candidate genes on chromosome 7, and a GWAS using 259 rice accessions revealed significant single nucleotide polymorphisms (SNPs) near several known resistance genes including xa25, Xa26, Xa21, Xa23, Xa33, xa13, and Os-11N3 genes. Dilla-Ermita et al. [17] conducted a GWAS on 285 rice accessions and identified novel SNPs linked to known Xa genes. The results found that five genes (xa5, Xa7, xa13, xa24, and xa34) originate from the aus genotype, while ten genes (Xa2, Xa4, Xa11, Xa14, Xa16, Xa18, xa25, Xa26, xa28, and Xa39) are derived from indica. Furthermore, a novel BLB resistance gene was identified on chromosome 11, designated as Xa43(t) within the 27.83–27.95 Mbp region by Kim and Reinke [23]. Moreover, Korinsak et al. [18] identified 147 genes on chromosome 1–6, 8, 9, and 11 using GWAS by phenotypic testing against 20 Xoo isolates. Among the identified genes, ten genes on chromosome 11 were found to be located close to well-known genes within the 17–29 Mb region.

Genotyping-by-sequencing (GBS) is a well-established method for generating genotypic information for GWAS. This technique provides a large number of SNPs for species diversity classification, haplotype map construction, the performance of GWAS, and the discovery of candidate SNPs [24,25,26]. Due to its effectiveness and low cost, this technique has been successfully incorporated into rice using restriction enzymes as the primary method for reducing the complexity of DNA samples and generating high-quality markers [25,27,28]. Many research studies of rice have incorporated this technique into their GWAS methods [17,29,30,31,32]. GWAS has become a powerful approach for unraveling the molecular genetics associated with natural phenotypic variation in rice. Various traits have been investigated using this technique, including agronomic traits, quality, and resistance to biotic and abiotic stresses [26,33,34,35,36]. Data collected from a variety of populations, plant growth stages, and different Xoo variations have been utilized in GWAS for gene mining, QTL mapping, and searches for candidate SNPs for BLB resistant traits in rice. In 2021, Shu et al. [25] conducted 259 rice lines including 240 indica and 19 japonica rice lines against two Xoo races (P3 and P6 races) and then discovered 2 NLR protein-encoding genes that exhibited a significantly high expression in the rice resistance lines. In 2022, Lu et al. [37] identified 13 quantitative trait loci (QTL) associated with BLB resistance (eight novel QTLs and five QTLs harboring known genes) located on chromosomes 5 and 11.

According to the World Agricultural Outlook Board’s report on rice from the US Department of Agriculture, Thailand ranks sixth worldwide in rice production and second in rice exports [38,39]. In 2018–2022, Thailand exported more than 35 million tons of rice [40]. In this regard, Thailand is one of the principal production regions for rice in the world. It has abundant rice genetic resources and a diversity of ecologies such as irrigated, rain-fed lowland, deep water, and upland ecosystems [41,42]. Thailand’s rice germplasm consists of improved cultivars, landraces, and local cultivars [43]. However, only one report has described the identification of a BLB resistance gene in Thailand’s germplasm using GWAS [18]. Therefore, the objective of this study is to identify candidate SNPs for BLB resistance genes in 200 accessions of Thailand’s local indica rice at the tillering stage against Thai Xoo isolates. To achieve this, we conducted a GWAS analysis and phenotypic evaluation by analyzing the average lesion length of rice panels against seven Thai Xoo isolates, along with genotyping data from 25,338 SNPs using the GBS method.

2. Materials and Methods

2.1. Plant Materials

Two hundred Thai landrace rice accessions were utilized to represent the diverse geographical regions and ecosystems found in Thailand. This comprehensive representation of rice diversity is crucial for breeding programs, genetic studies, and conservation efforts. The accessions included lowland, upland, and floating varieties, sourced from various regions of Thailand, with 70 from the northern region, 39 from the central region, 61 from the northeastern region, four from the western region, 14 from the eastern region, and six from the southern region. To grow the accessions for evaluation, the study was conducted under greenhouse conditions at the Thailand Rice Science Institute (TRSI) from October 2021 to January 2022, where the temperature ranged from 24 to 35 °C during the day and from 17 to 33 °C at night. The seed accessions were germinated on tissue paper for 48 h, and one week later were transplanted into pots containing three plants each, for a total of six plants per accession. Fertilization of the plants occurred at 15 and 21 days of age, using a 15-15-15 fertilizer and 46-0-0 fertilizer, respectively [44,45]. In addition to the study accessions, five standard rice varieties known for their resistance to the disease (DV85, IRBB5, IRBB21, RD7, and RD31) were included, as well as three susceptible rice varieties (TN1, IR24, and KDML105) as a comparison for the evaluation of disease resistance.

2.2. Evaluation of Bacterial Leaf Blight Resistance in Rice Plant

Seven Xoo isolates were selected for artificial inoculation of 200 Thai landrace rice accessions. The isolates are named SK2-3-F, XORE1-1, XONS2-1, 3XOBR2-2, 59XOCRCS7-3, 61XOSPSJ1-10, and 61XOSPSJ2-10 They were selected from a diversity of genetic and pathotype representations of Xoo. The Xoo cultures were multiplied for 48 h at 28 °C on nutrient agar medium and then suspended in sterilized water to a final concentration of 10⁹ CFU/mL. Each rice accession was inoculated with six plants using the leaf-clipping method at the tillering stage, with four fully expanded leaves clipped [46,47,48,49]. The lesion length (LL) was measured 14 days after inoculation or until a stable lesion was visible on the susceptible standard check variety [17,50]. The LL measurement was taken from the cutting edge of the leaf to the necrotic blight lesion in the middle vein of the leaf with the unit in centimeters [48]. The scoring was based on the standard method of the International Rice Research Institute (IRRI) [17,50,51,52,53]. Four groups of scoring were established based on LL, with lesion lengths ranging from 0–5 cm considered resistant, >5–10 cm moderately resistant, >10–15 cm moderately susceptible, and more than 15 cm susceptible [17,50,54]. The average value of lesion length was used as the phenotypic data for the GWAS analysis.

2.3. Genotyping Preparation

The study utilized two hundred BAM files, provided by the Ubon Ratchathani Rice Research Center in the Rice Department of Thailand, to obtain genotypic information on the rice varieties involved. The DNA library was generated using a ApeKI restriction enzyme, followed by ligation with adaptors specific to the Ion S5^TM XL Sequence (Thermo Fisher Scientific, Waltham, MA, USA). Using E-Gel^TM SizeSelect^TM agarose gels (Invitrogen, Waltham, MA, USA), 500–300 base pairs of fragments were selected. All sequence data were aligned to the Nipponbare genome as a reference genome (Nipponbare Reference-IRGSP-1.0) using the Ion Torrent^TM Suite Software Alignment Plugin version 5.2.2 (Thermo Fisher Scientific, Waltham, MA, USA) and then BAM files were automatically generated. Duplicate reads in the BAM format were removed using Picard Tools version 2.27.0 [55]. The BAM files were created as fastq files using Samtools v1.9 [56] and realigned with the japonica reference genome using a Burrow–Wheeler Aligner (BWA) v 0.7.17 [57] and SAMtools. Variant were called using the Genome Analysis Toolkit (GATK) software v 4.2.3.0 [42,58] where SNPs and insertions-deletions (indels) were named using the GVCF analysis, and variants were grouped using GATK’s CombineGVCFs function [59]. The joint call step was then performed using genotypeGVCFs of the GATK software [60], followed by SNP extraction using SelectVariantsGATK based on the reference genome. To improve the quality of the SNPs, Tassel v.5 was utilized to remove low quality SNPs that possessed a minor allele frequency less than 5% and a call rate below 75% [29,61,62,63]. Genetic datasets generated using GBS technology may contain missing values in SNP data due to low coverage or other technical issues. Imputation is a common method used to impute these missing values based on the linkage disequilibrium between SNPs in the dataset. Beagle version 5.4 is a widely used software tool for imputing genotypes in such datasets and is well-suited for use with GBS data. Therefore, in this study, we used Beagle version 5.4 to impute the missing values in the SNP data obtained from GBS genotyping. The algorithm is consistently high in accuracy within 90–100% for the major homozygous genotype, even with increasing missing rates [64,65,66]. As a result of filtering, a total of 25,338 high-quality SNPs were included in the GWAS analysis.

2.4. Population Genetics Structure Analysis, Principal Component Analysis, and Linkage Disequilibrium Assessment

The estimation of subgroup numbers in the rice panel was accomplished through the application of population structure analysis and principal component analysis (PCA). Following the imputation and pruning of data obtained from the panel of 200 rice accessions, 25,338 SNPs with a minor allele frequency greater than 5% were obtained. The STRUCTURE 2.3.4 program was utilized for estimating the population structure and K values were determined by setting the range from 1 to 10 to find the lowest coefficient of variation [28,67,68]. Both PCA and linkage disequilibrium (LD) analysis were performed using the Tassel [61]. The PCA was conducted on 25,338 SNPs obtained from a total of 4,086,284 SNPs within the rice panel through an initial analysis of all 200 rice accessions. Subpopulation verification was performed by combining and naming all the 200 BAM files with a representative japonica rice, followed by plotting using the RStudio program. Using the 25,338 highly accurate SNPs, Hapmap files were generated through the Tassel software. The LD was determined by calculating the pairwise squared correlation (r²) between SNP markers for each pair of loci. The LD analysis was conducted following the method described by Hill and Weir [69], based on a sliding window of 50 SNPs with a minor allele frequency of 0.05. Subsequently, a scatter plot with smoothing spline regression lines was developed. In addition to these values, the LD decay rate was estimated based on the physical distance between markers and the r² values and was determined when the r² value declined to half its average maximum value [70].

2.5. Genome-Wide Association Analysis

In this GWAS, multiple statistical methods were used to control for false-negative and false-positive results [71,72]. The Genomic Association and Prediction Integrated (GAPIT) tool was employed to perform the analysis, which was run using the R version 4.1.2 [73,74]. The 25,338 SNP loci were analyzed in relation to seven Xoo isolates to identify the association with BLB resistance. The four statistic methods were used as the mixed linear models (MLM) [75], Settlement of MLM Under Progressively Exclusive Relationship (SUPER) [76], fixed and random model circulation probability unification (FarmCPU) [77], and the Bayesian-information and linkage disequilibrium iteratively nested keyway (BLINK) method [78] were used to control for false-negative results [71]. The threshold for identifying significant SNPs was determined using the False Discovery Rate (FDR) [18,25,36,79,80]. Gene annotations within a 100 kb region surrounding the highly associated SNPs were taken into consideration for SNP identification. To visualize the results of the GWAS, Manhattan plots and quantile-quantile (Q-Q) plots were generated using the GAPIT tool in R. The scatter plot of the Manhattan plot provides a summary of the results by plotting the negative logarithm of the p-value obtained from each GWAS model against the genomic region of each SNP on the 12 rice chromosomes. On the other hand, the Q-Q plot visualizes the observed distribution of the negative logarithm of the p-values in comparison to their expected distribution under the null hypothesis.

2.6. Candidate SNP Identification

In order to explore candidate genes underlying BLB resistance, SNPs with a p-value less than 0.05 were considered as genomic regions carrying candidate genes. All gene loci within ±100 kb of the flanking regions of each SNP were extracted from the annotation of the Oryza sativa reference sequence (OsNippon-bare-Reference-IRGSP-1.0) [81]. To validate the results of GWAS, all annotated genes contained within the genomic regions were compared with genes known to be related to the phenotypic traits analyzed, as available in the Oryzabase database, the rice genome annotation project, and The Rice Annotation Project Database [20,82,83,84,85], or present in the literature as QTLs or genes. The Haploview 4.2 software was used to estimate the local LD block within ±100 kb of the genomic region containing significant SNPs [86]. Subsequent to identifying the candidate genes, haplotype analysis was conducted using all SNPs within their coding regions. A major haplotype was determined to be present when a minimum of 10 accessions were detected [37].

3. Results

3.1. Phenotypic Evaluation among Rice Panel

Through the inoculation of 200 rice varieties with seven Xoo isolates, distinct symptoms were identified between the various rice accessions. The lesion length distribution among Xoo isolates SK2-3-F, XORE1-1, XONS2-1, 3XOBR2-2, 59XOCRCS7-3, and 61XOSPSJ1-10 are presented in the bar graph (Figure 1a–g). The lesion length distribution is continuous and ranges from 1 to 35 cm. Thirty-one percent of 200 Thai landrace rice accessions showed a resistant reaction to Xoo isolate 59XOCRCS7-3, which was the highest percentage among the seven Xoo isolates used in this study (Figure 1h). While Xoo isolate SK2-3-F had 20% of the examined rice accessions, which showed a susceptible reaction.

Most rice accessions showed moderate resistance, as evidenced by a slight left skew distribution of the bar graphs. However, two accessions, namely ARDA-174 and ARDA 262, demonstrated broad-spectrum resistance to all Xoo isolates. The proportion of resistant to susceptible reactions were calculated and are shown in Figure 1h. Notably, the symptoms observed between the standard checks for resistance and susceptibility were significantly distinguishable (Figure 1i). The results showed a range of highly resistant reactions (HR: 0–5 cm), from 15.5% to 31%, while the range of susceptible reactions (moderately susceptible (MS) and highly susceptible (HS): ≥10) varied from 26% to 46.5% (Figure 1).

3.2. Genotypic Variation and Population Structure in Rice Panel

Imputation and error correction of the GBS data were performed. The VCF file of 4,086,620 SNPs from 200 Thai rice accessions was subjected to quality control procedures using Tassel version 5. SNPs with MAF less than 0.05% and a coverage of less than 75% were filtered out (Figure 2). These results suggested that imputation was an essential step in improving the quality of GBS data, and the use of Beagle 5.4 can be an effective method for imputing missing genotypic data. Additionally, heterozygosity alleles and indel sequences were removed before imputation to reduce the missing data rate [46]. The GWAS was conducted using 25,338 high-quality SNPs, which were distributed evenly across the 12 chromosomes. Chromosome 1 had the largest number of SNP loci, while chromosome 9 had the smallest. This result was consistent with the research of Lu et al. [37], who conducted GWAS of 421 rice accessions derived from the 3000 rice genome project (3KRGP), and the results showed that the highest and lowest number of SNPs were identified on chromosomes 1 and 9, respectively.

The analysis of 25,338 high-quality SNPs revealed two subpopulations in the rice panel, as supported by the estimation of the K value from the DeltaK plot generated by STRUCTURE 2.3.4 (Figure 3). A DeltaK plot identifies possible subpopulations in a dataset by detecting the peak in log-likelihood values at different values of K. In this study, the DeltaK plot revealed that the highest peaks were observed at K = 2 and a second peak at K = 6 (Figure 3), suggesting the presence of two or six distinct clusters within the 200 Thai rice accession panel. These results were consistent with a previous investigation [87,88], which classified the indica group into two major subpopulations (K = 2), but which reported four subgroups within those two subpopulations (K = 4).

The majority (86.8%) of pairwise kinship coefficients were found to be within the range of 0 to 0.5, while a minority (approximately 13.19%) of kinship values exceeded 0.5 (Figure 4a). These results indicated low levels of relatedness within the rice indica panel [87]. The eigenvalues associated with each PC represent the amount of variation explained by that component. In Figure 4b, we presented the eigenvalue plot obtained from the PCA. The plot showed a clear elbow point at PC = 2, indicating that the first two PCs capture most of the variance in the data. This suggested that the data can be effectively summarized and analyzed using only the first two PCs. Therefore, we retained the first two PCs for further analysis. A PCA was conducted to explore the genetic diversity and structure within the panel of rice accessions. The first two PCs (PC1 and PC2) were visualized to reveal the genetic variation among the accessions. However, the PCA plot indicated an ambiguous subpopulation structure (Figure 4c). PC1 and PC2 jointly accounted for roughly 14% of the total genetic variation, with PC1 and PC2 accounting for 8.5% and 5.5%, respectively. The two principal components captured some of the genetic variation among the accessions, but other factors may contribute to the overall genetic diversity within the panel. The findings demonstrated that genetic differentiation among the six geographic populations was not substantial, which may be attributed to genetic exchange between populations [88]. Based on the geographical origin of the rice accessions, the 200 rice accessions were separated along PC1, with accessions from central and eastern regions exhibiting negative values, while those from northeastern and northern regions had positive values. These findings agree with previous reports suggesting that indica rice varieties tend to cluster according to their geographical location [42,88,89].

To confirm the sub-species of rice, a panel of rice samples was utilized to perform SNP recall for 31,300 loci combined to 10 representative SNPs selected for the japonica rice group. Figure 4d shows a distinct separation between the indica and ten japonica rice groups. Out of the 200 rice samples analyzed, 18 were categorized as belonging to the admixture group, while only one sample was identified as being of the japonica group. The genetic variation in the sample population is geographically dependent, with the majority of the samples from the north and northeast located on the positive side of PC2, while the rest were located on the negative side of PC2. This suggests that most of the samples in the study are indica rice and admixture groups [42].

The current investigation estimated LD decay by evaluating 25,338 SNPs across a genomic region spanning 12 chromosomes. The findings revealed that the maximum r² value for the entire genome drops by half when the distance between SNPs reaches 118,257 base pairs, as depicted in Figure 5. The shortest LD decay distance was 72 kb, observed on chromosome 11, while the longest distance of LD decay is 225.8 kb, identified on chromosome 10, as indicated. These outcomes corresponded to prior studies, which suggested that the average LD decay distance for Indica rice is approximately 100 kb [18,70,87,90]. Since most marker distances are below 100 kb, it is expected that the study would have reasonable power to identify common large-effect variations in the Indica panel.

3.3. GWAS Analysis of Bacterial Leaf Blight Resistance Traits

Various statistical models were used to identify SNPs associated with BLB disease caused by Xoo in rice. The results showed statistical significance at a threshold of 1.97 × 10⁻⁶ and a false discovery rate (FDR) cutoff at 0.05. The MLM model, which considers fixed and random effects, identified four SNPs significantly associated with the XORE1-1 isolate on chromosome 2. Meanwhile, the SUPER method designed to control false positives detected 15 SNPs linked to SK2-3-F, XORE1-1, and 59XOCRCS7-3 isolates on chromosomes 1, 2, 4, and 7. Similarly, multi-locus models such as FarmCPU and BLINK identified 11 and 10 significant SNPs linked to SK2-3-F and 59XOCRCS7-3 isolates, respectively (

Table 1. The significant SNPs of 200 Thai rice accessions against 7 Xoo isolates using four statistical models in GWAS.

Method/Traits	SNP	Chr.	Allele	Pos	p-Value	MAF	H&B. p-Value	q-Value
SUPER.Xoo1	S1_27143044	1	A/C	27,143,044	1.97 × 10−6	0.26531	0.00919	5.71
SUPER.Xoo1	S1_27155997	1	A/G	27,155,997	7.17 × 10−7	0.26020	0.00454	6.14
SUPER.Xoo1	S7_3529421	7	A/G	3,529,421	4.70 × 10−7	0.15816	0.00397	6.33
SUPER.Xoo1	S7_3537960	7	T/C	3,537,960	4.70 × 10−7	0.15816	0.00397	6.33
SUPER.Xoo1	S7_18664455	7	A/G	18,664,455	4.47 × 10−7	0.21939	0.00397	6.35
FarmCPU.Xoo1	S1_27155997	1	A/G	27,155,997	3.52 × 10−8	0.26020	0.00083	7.45
FarmCPU.Xoo1	S2_935021	2	A/T	935,021	1.34 × 10−7	0.32653	0.00113	6.87
FarmCPU.Xoo1	S5_18901969	5	C/A	18,901,969	1.64 × 10−6	0.13776	0.01036	5.79
FarmCPU.Xoo1	S7_3529421	7	A/G	3,529,421	6.57 × 10−8	0.15816	0.00083	7.18
FarmCPU.Xoo1	S8_6216184	8	A/G	6,216,184	4.19 × 10−6	0.46429	0.02126	5.38
FarmCPU.Xoo1	S4_28336644	4	C/T	28,336,644	1.22 × 10−5	0.18878	0.05167	4.91
FarmCPU.Xoo1	S2_23903211	2	G/T	23,903,211	1.44 × 10−5	0.10459	0.05223	4.84
BLINK.Xoo1	S2_935021	2	A/T	935,021	5.78 × 10−8	0.32653	0.00073	7.24
BLINK.Xoo1	S7_3529421	7	A/G	3,529,421	6.07 × 10−10	0.15816	0.00002	9.22
BLINK.Xoo1	S7_16039418	7	A/T	16,039,418	7.35 × 10−7	0.29592	0.00465	6.13
BLINK.Xoo1	S8_6216184	8	A/G	6,216,184	2.57 × 10−7	0.46429	0.00217	6.59
MLM.Xoo2	S2_25400454	2	G/A	25,400,454	1.09 × 10−6	0.43333	0.01181	5.96
MLM.Xoo2	S2_25408622	2	G/A	25,408,622	1.77 × 10−6	0.43846	0.01181	5.75
MLM.Xoo2	S2_25413622	2	G/A	25,413,622	1.77 × 10−6	0.43846	0.01181	5.75
MLM.Xoo2	S2_25400408	2	T/C	25,400,408	2.80 × 10−6	0.43590	0.01181	5.55
MLM.Xoo2	S2_25400480	2	A/C	25,400,480	2.80 × 10−6	0.43590	0.01181	5.55
MLM.Xoo2	S2_25400530	2	A/C	25,400,530	2.80 × 10−6	0.43590	0.01181	5.55
MLM.Xoo2	S2_25400580	2	C/T	25,400,580	6.24 × 10−6	0.43077	0.02260	5.20
SUPER.Xoo2	S2_25400454	2	G/A	25,400,454	4.77 × 10−6	0.43333	0.03765	5.32
SUPER.Xoo2	S2_25408622	2	G/A	25,408,622	6.73 × 10−6	0.43846	0.03765	5.17
SUPER.Xoo2	S2_25413622	2	G/A	25,413,622	6.73 × 10−6	0.43846	0.03765	5.17
SUPER.Xoo2	S2_25400408	2	T/C	25,400,408	8.92 × 10−6	0.43590	0.03765	5.05
SUPER.Xoo2	S2_25400480	2	A/C	25,400,480	8.92 × 10−6	0.43590	0.03765	5.05
SUPER.Xoo2	S2_25400530	2	A/C	25,400,530	8.92 × 10−6	0.43590	0.03765	5.05
SUPER.Xoo2	S4_31500482	4	C/G	31,500,482	1.43 × 10−5	0.45385	0.04039	4.84
SUPER.Xoo2	S4_31505231	4	T/G	31,505,231	1.43 × 10−5	0.45385	0.04039	4.84
SUPER.Xoo2	S4_23423399	4	G/A	23,423,399	1.43 × 10−5	0.46667	0.04039	4.84
BLINK.Xoo2	S2_25400454	2	G/A	25,400,454	3.96 × 10−12	0.43333	0.00000	11.40
BLINK.Xoo2	S4_31770194	4	C/G	31,770,194	7.04 × 10−8	0.46410	0.00089	7.15
SUPER.Xoo5	S1_27155997	1	A/G	27,155,997	9.82 × 10−8	0.36592	0.00249	7.01
SUPER.Xoo5	S1_27143044	1	A/C	27,143,044	2.55 × 10−7	0.36313	0.00322	6.59
FarmCPU.Xoo5	S12_15567690	12	G/A	15,567,690	3.45 × 10−9	0.39385	0.00009	8.46
FarmCPU.Xoo5	S1_22492602	1	G/A	22,492,602	6.15 × 10−8	0.45810	0.00078	7.21
FarmCPU.Xoo5	S3_438061	3	A/G	438,061	6.39 × 10−7	0.43017	0.00540	6.19
FarmCPU.Xoo5	S5_2313550	5	G/T	2,313,550	2.61 × 10−6	0.39944	0.01656	5.58
FarmCPU.Xoo5	S6_1362133	6	C/T	1,362,133	3.53 × 10−6	0.37989	0.01789	5.45
FarmCPU.Xoo5	S12_18758444	12	T/C	18,758,444	8.17 × 10−6	0.39944	0.03042	5.09
FarmCPU.Xoo5	S1_30147885	1	T/C	30,147,885	8.40 × 10−6	0.37709	0.03042	5.08
BLINK.Xoo5	S1_27155997	1	A/G	27,155,997	4.55 × 10−11	0.36592	0.00000	10.30
BLINK.Xoo5	S8_26134412	8	C/A	26,134,412	7.65 × 10−8	0.42737	0.00097	7.12
BLINK.Xoo5	S2_25453856	2	A/C	25,453,856	5.97 × 10−7	0.44413	0.00504	6.22
BLINK.Xoo5	S2_18882825	2	A/G	18,882,825	5.00 × 10−6	0.45810	0.03167	5.30

). Notably, all models discovered significant SNPs associated with only three Xoo isolates (SK2-3-F, XORE1-1, and 59XOCRCS7-3), emphasizing the significance of isolates-specific effects in identifying SNP-trait associations. Overall, the findings contributed insights into the genetic basis of BLB resistance in rice and demonstrate the usefulness of different statistical models in SNP-trait association studies.

Based on the Q-Q plots, the MLM, FarmCPU, and BLINK models exhibited a linear pattern with a distinct upward deviation in the tail region, suggesting effective control of both false positives and false negatives. However, MLM appeared to have a limited ability to detect the associated SNPs. On the other hand, the SUPER model displayed a non-normal distribution on the lower left corner, which might be indicative of a ceiling effect or right-skewness in the data. Through GWAS, we identified significant SNPs in three Xoo isolates using both SUPER and BLINK analysis. However, we found that the Q-Q plots of FarmCPU and BLINK provided a better fit to the data, as the plotted points followed a straight line closely along the identity line with a sharply deviated tail (Figure 6, Figure 7 and Figure 8). This distinctive pattern was more apparent than that observed in other models. The candidate genes containing SNPs significantly associated with bacterial leaf blight resistance trait were identified by the gene annotation [81] and the Oryzabase database [82]. In summary, GWAS revealed 32 significant SNPs associated with BLB resistance, which were located on chromosomes 1–8 and 12. The SNP with the highest level of significance was identified at S7_3529421 on chromosome 7, with a p-value of 6.07 × 10⁻¹⁰. This result was obtained using the BLINK method against SK2-3-F, and the details are provided in Table 1.

The Manhattan plots showed a set of highly significant SNPs that were associated with the SK2-3-F isolate, as identified by four distinct methods (Figure 6). These results were generated using the SUPER model with the genomic coordinates S1_2714304, S1_2715599, S7_3529421, S7_3537960, and S7_1866445 on chromosomes 1 and 7. The corresponding −log₁₀ (p-values) for these SNPs ranged from 5.70 to 6.34. Notably, the FarmCPU model generated the largest number of significant SNPs, spanning across chromosomes 1, 2, 4, 5, 7, and 8, with −log₁₀ (p-values) ranging from 4.84 to 7.45 at FDR threshold of 0.05. The BLINK model identified a limited number of significant SNPs associated with resistance to the SK2-3-F isolate. Specifically, this model identified four significant SNPs located on chromosomes 2, 7, and 8. However, it demonstrated the highest peak of significance among all methods, with a −log₁₀ (p-value) of 9.22.

Manhattan plot showed the significant SNPs associated with BLB inoculated with XORE1-1 isolate (Figure 7). The findings revealed that only two out of the four statistical models generated significant SNPs associated with the BLB resistance gene. The MLM models produced the highest number of significant SNPs, with most located on chromosome 2. In contrast, the BLINK models only identified two significant SNPs on chromosomes 2 and 4. In addition, the Manhattan plot showed the significant SNPs linked to BLB resistance against the 59XOCRCS7-3 isolate (Figure 8). The results generated from the SUPER, FarmCPU, and BLINK models showed highly significant SNPs along chromosomes 1, 2, 3, 5, 6, 8, and 12. These findings indicated that different statistical models can reveal significant SNPs associated with the resistance to BLB caused by different Xoo strains. Therefore, the use of multiple models can provide a more comprehensive understanding of the genetic factors underlying BLB resistance in rice. Notably, both the BLINK and SUPER models were able to detect significant SNPs in all three Xoo isolates (SK2-3-F, XORE1-1, and 59XOCRCS7-3). Among the four models, BLINK method performed the best in controlling false positive and false negative errors, as evidenced by an analysis of the Q-Q plots, which is a common approach to assess model performance. The Q-Q plot generated by the SUPER model showed an inflated upward line, indicating a high number of false positives. In contrast, the BLINK model produced a straight line with a slight tail, suggesting that it effectively controlled both false positives and false negatives.

Out of the 581 genes that were located surrounding the significant SNPs, 179 genes were found within the ±100 kb region surrounding the SNP of interest. Among these 179 genes, 49 were selected as candidate genes based on their known functions in plant defense mechanisms. In addition, they were selected for their proximity to significant SNPs in the gene region. This study focuses on genes associated with several plant defense functions. These genes include the serine/threonine protein kinase-related domain-containing protein, the WRKY transcription factor, the cell cycle-associated protein kinase, the Leucine-rich repeat, the NBS-LRR disease resistance protein, disease resistance protein RGA2, and the MYB transcription factor.

3.4. Analysis of Haplotypes and Linkage of Candidate Genes

In order to perform a haplotype analysis to detect significant differences between the resistance and susceptibility, 16 candidate SNP loci were selected based on the gene annotation as a plant defense mechanism located on chromosome 2, 4, and 8. One of the significant SNP loci, S2_25453856, had two candidate genes surrounding its region (±100 kb), namely LOC_Os02g42310.1 and LOC_Os02g42314.2, which were located within the LD decay. These genes started from the position 25442875 to 25450093 and 25450313 to 25454074, respectively.

The LOC_Os02g42310.1 gene is composed of 6 SNPs with 5 haplotypes, and the function of the gene is predicted to be a serine carboxypeptidase II-like protein. Serine carboxypeptidases are a group of enzymes that are involved in cleaving C-terminal amino acids from peptide substrates. These enzymes have been shown to play important roles in various biological processes in plants, including cell wall metabolism and defense responses against pathogens. One specific serine carboxypeptidase gene has been identified as being significantly associated with sheath blight resistance [91]. Six SNPs identified within this gene have the potential to affect the activity or stability of the protein, which could, in turn, impact the plant’s resistance or susceptibility to the disease. These findings suggest that the identification and characterization of specific genetic markers, such as SNPs in this gene, could be a valuable tool for developing more effective strategies for breeding rice varieties with enhanced disease resistance. The other gene locus, LOC_Os02g42314.2, is composed of three SNPs that can produce seven different haplotypes (Hap1-Hap7). The protein encoded by this gene is a Ubiquitin-conjugating enzyme/RWD-like domain-containing protein.

Moreover, two gene loci in the same linkage disequilibrium decay with S8_6216184 on chromosome 8 were identified, LOC_Os08g10440.1, which encodes a CC-NBS-LRR protein known to confer Pyricularia oryzae resistance-33(t) in rice [92]. This gene, located on chromosome 8, consists of seven different genetic patterns. Interestingly, haplotypes Hap7 and Hap8 of this gene were found to confer a moderated resistance (MR) phenotype, where the lesion length (LL) was greater than or equal to 10 cm. LOC_Os08g10560.1 is responsible for encoding the nuclear factor Y C subunit 5 (NF-YC5) in rice, which plays a crucial role in the processes of DNA replication and repair. The gene ontology annotation suggested that the OsNF-Y genes are co-expressed with stress response, seed storage reserve accumulation, and plant development genes [93]. NF-Y is a heterotrimeric transcription factor that regulates a variety of biological processes in plants, including defense mechanisms against both biotic and abiotic stresses. In particular, the NF-YC5 subunit of the NF-Y complex is known to regulate the plant defense response in various species, such as Arabidopsis thaliana and Cassava (Manihot esculenta). The expression of NF-YC5 is induced upon infection by a range of plant pathogens, including bacteria, fungi, and viruses. [94,95]. The gene comprises 13 SNPs, resulting in 8 haplotypes. Among these, Hap1, Hap4, and Hap9 demonstrated a moderated resistance reaction, as evidenced by a lesion length (LL) of 5 ≥ 10 cm.

LOC_Os04g52900 was associated with SNP S4_31500482 and S4_31505231, which encode for ATP binding cassette (ABC) transporters, a well-studied and highly conserved family of transporter proteins. Previous research has demonstrated that certain ABC transporters play a role in conferring resistance against bacterial pathogens, such as the HpaABC transporter which is involved in bacterial blight resistance [92,96]. In addition, the genetic locus LOC_Os04g52900 contains several SNP loci, resulting in a genetic variant with three distinct haplotypes (Figure 9). Of these haplotypes, Hap2 exhibited a significantly enhanced level of resistance, as evidenced by a lesion length of less than or equal to 5 cm. Additionally, our investigation suggested that genetic variation within the LOC_Os04g53350 gene, annotated as a conserved hypothetical protein, may also play a crucial role in providing disease resistance in plants. This gene demonstrated a highly resistant reaction (LL = 4 cm) against the SK2-3-F pathogen and is located on chromosome 4 within a gene region containing three genes. Our findings highlight the potential of these genes as targets for improving plant disease resistance through breeding programs.

4. Discussion

BLB, caused by Xoo bacteria, have reported yield losses ranging from 80–100% in Asia [4,5,12]. GBS is an advanced technique that has emerged as a valuable tool for identifying genetic variants linked to BLB resistance in crops through GWAS. Despite the challenges associated with GBS, it is still a cost-effective, high-throughput, and minimally invasive method, making it suitable for large-scale studies of diverse crop populations. GBS can capture rare and low-frequency genetic variants that may not be detected by other genotyping methods, providing a more comprehensive view of the genetic architecture underlying BLB resistance [29,60,97].

In this study, we explored the genetic basis of BLB resistance in Thai local rice. To do this, we conducted a GWAS on a panel of 200 indica rice accessions collected from six different regions in Thailand. We can classify the rice panel into two subpopulations, which is consistent with the previous report that used indica rice for genetic analysis [42,87]. We assessed BLB resistance by infecting all 200 rice accessions with seven Thai Xoo isolates. The four methods of statistical analysis were used for GWAS analysis. The Q-Q plots revealed a non-uniform distribution of p-values in the SUPER models of all empirical traits, consistent with previous research, indicating their unsuitability for trait-associated mapping due to the possibility of false marker-trait associations [72]. To correct population structure and family relatedness, complex models such as MLM, FarmCPU, and BLINK have been suggested [72,76,77,78,79]. MLM and FarmCPU Q-Q plots displayed a straight line with slightly deviated tails, indicating a decrease in false positives but an increase in false negatives due to overfitting, which is in line with other studies. However, the Q-Q plot of the multi-locus model, BLINK, indicated that it controls both false positives and false negatives for all empirical traits in both crops, as shown by a straight line with a sharp deviated tail. As a result of our GWAS analysis, in this study, we have identified a novel region that contains SNPs associated with three Xoo isolates. These SNPs were found to be located on chromosomes 1–8 and 12 but not on chromosomes 9–11. It is worth noting that previous reports have mostly identified the BLB resistance gene on chromosome 11 [17,18,36,98]. However, our results indicate that SNPs on chromosome 11 were only detected at a low level of significance, with a q-value threshold of 5.71 and an FDR of 0.01 for the three Xoo isolates. In addition, our investigation revealed 15 MSU genes flanking the well-known resistance gene Xa1, which were also located within the same LD block. Previous studies have reported that Xa2, Xa12, Xa14, Xa25, Xa31(t), and Xa38 were localized to the region near Xa1 and conferred race-specific resistance against Xoo [99]. Our finding indicated that the two significant SNPs, S4_31500482 and S4_31505231, consist of 26 MSU genes situated around them. Out of 15 MSU genes, LOC_Os04g52890, LOC_Os04g52870, and LOC_Os04g52900 have been identified as genes for a blast resistance gene Pikahei-1(t) mapped to chromosome 4 [17]. We also identified a gene, LOC_Os04g53160, which is located within the same LD decay as Xa1 and encodes BED-NLBs. It is a DNA-binding domain present in many transposases and chromatin-boundary element-binding proteins and acts as an integrated decoy domain. This gene was previously reported as an R gene analog named RGAg (Os04g53160) [100]. Additionally, we found that LOC_Os04g53160 was located near LOC_Os04g53170 and LOC_Os04g53180, which were identified as a putative Xa1 gene and a bromodomain-containing protein, respectively [100]. These findings provide valuable insights into the potential roles of these genes in BLB resistance and suggest new avenues for further exploration in developing BLB-resistant rice varieties. Furthermore, we found that this region contains LOC_Os04g53210 and LOC_Os04g53214, which encode for a glycolate oxidase (GLO) enzyme. Previous studies have reported that a reduction in the expression of GLO can lead to an increase in the production of reactive oxygen species (ROS), which can ultimately enhance the plant’s ability to defend against various stresses [101]. In addition, this study has identified seven NBS-LRR coding genes (LOC_Os02g02660.1, LOC_Os02g02670.1, LOC_Os02g39660.1, LOC_Os04g53050.1, LOC_Os06g03500.1, LOC_Os07g31500.1, and LOC_Os08g10440.1) located on chromosomes 2, 4, 6, 7, and 8, which may be crucial for the recognition and constant surveillance of invading elicitors and play a significant role in the plant defense mechanism [72,86,102]. Among these genes, LOC_Os04g53050 was identified as an Xa38 gene associated with S4_31500482 and S4_31505231 [4]. In a previous study, Dilla-Ermita et al. [17] identified eight SNPs located on chromosome 8 that were adjacent to and flanked the xa13 gene, including S8_26013849, S8_26044514, S8_26044528, S8_27319974, S8_27520607, S8_27536844, S8_27641374, and S8_27663433. Interestingly, these SNPs also flank our S8_26134412, which is located on the LOC_Os08g41380 gene that encodes a P-type PPR protein and has been implicated in responding to biotic and abiotic stresses [103]. Shu et al. [25] also reported the identification of S12_175769641 and S8_27163888, which are located near the clone R gene xa25 and Xa23, respectively. Interestingly, these two significant SNPs are located in close proximity to our own SNP results.

The SNPs with the highest level of significance associated with BLB resistance are located on chromosome 2 (Table 1) with a p-value of 3.69 × 10⁻¹². Among these, the candidate SNP S2_25400454 is located on non-coding regions [84], which are linked to several genes including LOC_Os02g42110 (cell cycle-associated protein kinase), LOC_Os02g42150 (wall-associated kinase, which positively regulates rice blast resistance), LOC_Os02g42200 (BB-block binding subunit of TFIIIC domain containing protein), and LOC_Os02g42310 (similar to serine carboxypeptidase II-like protein), all of which are in the same LD decay block (± 100 kb from the significant SNP). Finally, our analysis of flashpoint regions revealed that the LOC_Os04g53350 gene contains six haplotypes, among which Hap4 (CCGCAGC) showed a strong resistance reaction. This gene is located on chromosome 4 within a gene region that contains three other genes. Our findings suggest that LOC_Os04g53350 may play a crucial role in providing disease resistance in plants and could be a promising target for improving plant disease resistance through breeding programs.

The identification of a candidate gene in this study highlights its potential role in regulating immunity to BLB resistance in rice. The results provide valuable insights into the underlying genetic mechanisms of BLB resistance, which can help to improve breeding programs. Further investigations into the specific roles of these genes in enhancing rice’s resistance to this devastating disease are essential for the development of effective management strategies. These findings underscore the importance of continued research in this area to ensure the sustainability of rice production and food security worldwide.

5. Conclusions

The Genome-Wide Association Study (GWAS) approach was adopted to identify Single Nucleotide Polymorphisms (SNPs) associated with resistance to each of the seven Xoo isolates. To analyze the results, Manhattan and Q-Q plots were used, and it was found that the BLINK method outperformed other methods in terms of smoother test lines on the Q-Q plots. The study also identified candidate genes that could potentially confer resistance against Xoo. Two genes, LOC_Os04g53350 and LOC_Os04g52900, were found to exhibit highly resistant action against Xoo, with lesion lengths of 4 and 5 cm (Hap4 and Hap2 of each gene), respectively. Additionally, ten candidate genes were analyzed for haplotypes. The results of this research could be used to develop more efficient and effective breeding programs aimed at enhancing the resistance of rice against Xoo, ultimately leading to improved crop productivity and food security.