Genome-wide association analysis identifies resistance loci for bacterial blight in diverse East African rice germplasm

Xanthomonas oryzae pv. oryzae (Xoo), the causal agent of rice bacterial blight disease has been extensively characterized


INTRODUCTION
Bacterial blight (BB), caused by Xanthomonas oryzae pv. oryzae (Xoo), is one of the most devastating and economically important diseases of rice (Oryza sativa L.) all over the world (Savary et al., 2019). Rice resistance against BB can be generally divided into two main categories; the qualitative resistance controlled by major resistance (R) genes, and the quantitative resistance conferred by multiple minor genes or quantitative trait loci (QTLs) (Ramalingam et al., 2003;Deng et al., 2012;Bossa-Castro et al., 2018).
Although the disease resistance conferred by a single R gene is usually effective against certain races of the X. oryzae pv. oryzae pathogen, the resistance is easily breakdown due to greater selection pressure on pathogen evolution. Conversely, the quantitative resistance mediated by QTLs is presumably non-racespecific and is considered more durable (Liu et al., 2016). Thus, it has attracted more attention in the past decades and more than 70 QTLs for BB resistance have been identified Han et al., 2014;Djedatin et al., 2016;Dilla-Ermita et al., 2017;Zhang et al., 2017;Bossa-Castro et al., 2018).
Deployment of broad-spectrum resistant rice cultivars is considered the most effective and environmentally friendly way to control bacterial blight , quantitative resistance has been considered as a preferred strategy to achieve durable resistance although marker-assisted selection has not been effectively used for the improvement of BB resistance in rice. This issue is attributed to the polygenic nature of the trait and each QTL has a small effect. It is therefore difficult to accumulate multiple QTLs with small effects in breeding. In addition, most of the QTLs for BB resistance were identified and mapped using bi-parental population QTL analysis in the past decades. Because of limited molecular markers used and fewer recombinants in a primary mapping population, most of the QTLs for BB resistance are mapped to a region of 10 ~ 30 cM (Yang et al., 2021). Since a prerequisite for successful markerassisted selection (MAS) is the availability of markers that are closely linked with the target gene, the inaccuracy of Okello et al. 147 QTL mapping hinders the application of MAS. Therefore, the discovery of the large-effect QTLs and the use of a more powerful approach for the genetic dissection of complex traits are crucial to address this issue challenge in rice production. Genome-wide association studies (GWAS) present another alternative with two main advantages: (1) it can use a nature population instead of a bi-parental population. The rice varieties used in GWAS contain much more genetic diversity than the bi-parental lines used in segregation populations. Because using diverse germplasm for QTL mapping in GWAS, favors the identification of large-effect and novel QTLs; (2) most GWAS can result in a relatively high mapping resolution due to the existence of numerous historical recombination events (Takeda and Matsuoka, 2008) and using plenty of SNPs for association mapping. Therefore, GWAS provides a powerful tool for large-scale and precise identification of QTLs for complex traits like BB resistance in germplasm (Zhao et al., 2011;Han and Huang, 2013;Zhang et al., 2017;Zhai et al., 2018).
In this study, we identified the distribution of BB resistance genes of the East African accessions that are distributed and bred in the regions. We detected two loci on chromosomes 6 and 12 that carry the genes (Xa7, Xa27, Xa33) and (Xa25), respectively.

Plant materials
The selection of diverse panel breeding lines, varieties, and landraces including accessions of 78 genotypes was collected from accessions obtained from National Crops Resources Research Institute (NaCRRI), Namulonge, Uganda. The diverse germplasm collection was obtained from the East African countries of Uganda, Kenya, Tanzania, Rwanda, and Burundi. The field studies were carried out at Africa Rice Centre Mbe station Cote D'Ivoire in accordance with institutional, national, and international guidelines and legislation. The germplasm used in this research was non-GMOs.

BB resistance screening
The experiment was carried out in a contained screen house facility to prevent the spread of inoculum and was done using RCBD splitplot with nested subplot design. The four X. oryzae pv. oryzae strains were the main plots and, in each plot, a sub-plot of 78 genotypes was nested into early maturing and medium maturing Genotypes.
The BB resistance screening was performed 6 weeks after sowing. A total of four strains representing different races of X.  African and Philippine strains to specific Xoo races. b Naturally occurring Xa genes known to confer resistance against specific Xoo races based on a review. Source: Nino- Liu et al. (2006) andVerdier et al. (2011). oryzae pv. oryzae were used, namely: Xoo3.1, BAI3, MAI1, and PXO99.
Inoculation of the rice plants was done by cutting 1 to 2 cm of the leaf tip with a pair of scissors dipped in bacteria suspension (Kauffman et al., 1973). The screening was replicated two times over time. For each replicate, ten leaves of two plants per entry were inoculated with the X. oryzae pv. oryzae strains. Lesion lengths were measured 14 days after inoculation. Genotypes having lesion lengths ranging from 1 to 5 cm were rated resistant (R), 5 to 10 cm were rated as medium-resistant (MR), 10 to 15 cm were rated as medium-susceptible (MS), and those having greater than 15 cm were rated as susceptible (S).

Association mapping
The East African accessions comprising 78 germplasm samples were selected and phenotyped by measuring the lesion length of leaves inoculated with BB races Xoo3.1, BAI3, MAI1, and PXO99.

Linkage analysis and QTL mapping
TASSEL version 5.2 software (Bradbury et al., 2007) was used for manipulating and filtering SNPs for genome-wide association analysis. The initial dataset was filtered based on MAF 1 and 95% call rate. A unified mixed-model approach was deployed to account for population structure and familial relatedness Price et al., 2006). A compressed mixed linear model (MLM) was used to analyze association, considering population structure (Q) and relatedness or kinship (K) to reduce spurious associations .

BB resistance screening and population structure
Four diverse strains representing four different races of X. oryzae pv. oryzae (Table 1) were used to screen 78 genotypes. The lesion length (LL) distribution in the 78 accessions inoculated with four Xoo strains (PX099, MAI1, BAI3, and Xoo3.1) showed large phenotypic variation (Table 2). Among the inoculated cultivars, five accessions were highly resistant to all four strains with LL< 5 cm, and only highly susceptible to all four strains with LL≥15 cm (highlighted in white and red, respectively in (Figure 1)). Two accessions from Tanzania (Afaa Milela and Mbawa Mbili) conferred high resistance to all four strains from the Philippines (PX099), West Africa (MAI1 and BAI3), and the East African strain (Xoo3.1) Based on the LL of all the accessions, the four X. oryzae pv. oryzae strains were divided into three groups, namely the West African strain group (BAI3, MAI1), East African (Xoo,3.1), and Philippines strain group (PX099) (Figure 1). According to classification by LL, a large proportion of the accessions were resistant with 39.24, 35.44, 40.51 and 50.63% for strains Xoo3.1, BAI3, MAI1, and PX099 respectively (Figure 1).
The sources of resistant East African germplasm from different backgrounds can provide valuable material for facilitating breeding for BB resistance.

Identification of resistance loci against
To dissect genome-wide associated resistance loci for two West African X. oryzae pv. oryzae strains, one East African X. oryzae pv. oryzae strain, and one Philippines X. oryzae pv. oryzae strain, we performed GWAS with a mixed linear model of the TASSEL ver. 5.2 program, using 2,612 high-quality SNPs and LL as genotype and phenotype data, respectively (Figure 2). Based on the effective number of independent markers, the threshold of significant P-value was estimated to be 3.0E-5 by the Bonferroni correction method. In total, we identified 2 QTL within 1901 unique SNPs associated with BB resistance to four strains.

Significant resistance genes hotspots
GWAS results from this study indicated chromosomes 6 and 12 as hotspots for BB resistance. SNPs consistently associated with resistance against several strains were identified by GWAS. Known genes and loci (Xa) conferring resistance to specific Xoo races (Table 1) were overlaid with the Manhattan plots. Among the known Xa loci, the regions on chromosome 6 (Xa7, Xa27, Xa33(t)) and Xa25 on chromosome 12 (Wang et al., 2001;Blair et Chen et al., 2008;Chu et al., 2006a, b;Bao et al., 2010;Song et al., 1995;Chen et al., 2002) were identified as overlapping with significant SNPs in this study.

DISCUSSION
This study was conducted to screen and identify diseaseresistant rice cultivars, as well as key functional genes applicable for breeding new varieties with broadspectrum BB resistance. Bacterial blight is one of the most diseases of rice, causing significant yield losses in rice-growing ecologies throughout Africa (Kim, 2018). The current BB management strategies are not effective due to the rapid development of virulent Xoo strains (Kim et al., 2015). In Africa where rice bacterial blight outbreaks can be epidemic, new Xoo strains need to be identified and characterized.
In the present study, we identified two accessions from Tanzania (Afaa Milela and Mbawa Mbili) that are highly resistant to all four X. oryzae pv. oryzae strains (PX099, MAI1, BAI3, and Xoo3.1). The data generated in this study regarding these BB-resistant accessions will be utilized for rice breeding programs.
Recently, many R genes for BB were successfully incorporated into both elite inbred varieties and parental lines of hybrid rice to control the disease using MAS (Chukwu et al., 2019;Jiang et al., 2020;Li et al., 2020). Markedly, a few of those R genes, such as Xa3, Xa4, Xa7, and Xa21, have been widely utilized in rice Figure 2. Genome-wide association study of rice resistance to four Xoo strains. A, C, E, G Manhattan plots of GWAS results for strain BAI3, MAI1, PX099, and Xoo3.1, respectively. B, D, F, H Quantile plots of expected and observed -log 10 (P-value) for strains BAI3, MAI1, PX099, and Xoo3.1, respectively. The horizontal blue line indicates the significant P-value threshold of 3.0E-5. The arrow indicates the reported bacterial blight resistance genes.

Conclusion
The results from this GWAS have pinpointed resistance loci conferring differential resistance to the four representative strains of X. oryzae pv. oryzae in the East African accession. Two resistant loci were identified through this analysis. Using efficient phenotypic data and SNPs from genotyping is a powerful tool to decipher disease resistance in rice. The SNPs associated with X. oryzae pv. oryzae resistance would be helpful in the development of SNP markers for marker-assisted selection and tracking of known Xa genes. Effective monitoring of resistance genes in the breeding pipeline guides breeders on which varieties to deploy to specific areas depending on the Xoo population. The full potential of the novel loci identified loci in the East African germplasm will be unraveled through expression profiling. The genotypes that have exhibited a high degree of resistance but have no resistance alleles for specific Xa QTLs serve as new sources of resistance loci to plant breeders to diversify the genetic base of core breeding sets. Further analysis of the East African germplasm is recommended to identify potentially novel loci which were not detected in this study.