IDENTIFICATION OF IRON TOLERANT CANDIDATE LOCI IN RICE DETERMINED THROUGH GENOME-WIDE ASSOCIATION STUDY

Iron (Fe) toxicity is a significant abiotic stress in swamp land. The study aimed to identify the candidate loci related to Fe toxicity tolerance through Genome-Wide Association Study (GWAS) approach.  The study used 288 rice accessions consisting of 192 breeding lines and 50 local landraces, and custom-designed 384 rice SNPs-chips. A field evaluation was conducted in inland swamp for two season periods (2014 and 2015). Phenotypic data and association mapping were analyzed using XLSTAT and TASSEL 3.0. The candidate loci were analyzed by functional gene detection of the significant SNPs aligned to the Rice Annotation Project and the Institute for Genomic Research databases. Three linkage disequilibrium (LD) blocks were detected in the Fe tolerant population around the significant SNPs. The first LD block was mapped in chromosome 1 (the AtIRT gene and qFETOX1; qFETOX1-3 QTLs loci) resembled partitioning of Fe-toxicity tolerant mechanism. The second LD blocks located in chromosome 2 (qFE-TOX-2-1 and qFETOX-2 QTLs loci) and chromosome 3 (qFETOX-3 QTL, OsNAS1 and OsNAS2 loci), probably contributed to Fe exclusion mechanism. The third LD blocks located in chromosome 4 (OsFRO2 and qFETOX-4 QTL loci) and chromosome 7 (OsIRT2 and NAS3 loci). The third LD block found on tolerant genotypes both on vegetative and generative stages. This condition indicated that these loci were presumed playing a role for Fe toxicity tolerance in rice. Result of the study are beneficial for determining the strategy on developing Fe-toxicity tolerant rice for specific swamp land type through breeding programs.


INTRODUCTION
Ferrum/iron (Fe) is a micronutrient essential for plants because it plays a role in the metabolic processes such as DNA synthesis, respiration, and electron transport support photosynthesis process. Iron also acts as an electron acceptor in the redox reaction and activator for important enzymes in plant metabolism. Nevertheless, in acid soils, the soluble Fe could be available excessively (more than 300 mg kg -1 ) resulted in toxic effects to plants (Dobermann and Fairhurst 2000). Iron toxicity is one of the important abiotic stresses that can decrease rice production. Millions of hectares of rice fields in Asia, Africa, and Latin America were reported suffering iron toxicity (Matthus et al. 2015). In Indonesia, rice fields suffering from iron toxicity is spread on suboptimal soils, such as swamp area, tidal land, red-yellow podsolic land, lowland with poor drainage, and new crop areas scattered in many islands of Indonesia. The estimated hectarage of rice fields with a high content of Fe in Indonesia reached one million hectares (Suhartini 2004).
Development of iron-toxicity tolerant rice variety through breeding program seems to be a practical approach in dealing with iron toxicity stress in rice. In principle, physiological strategies can be targeted to address the iron toxicity problem. The strategies include: (1) Fe 2+ exclusion mechanism on the root surface through the Fe 2+ oxidation process into insoluble Fe 3+ . This strategy leads to the plaque formations on the root surfaces. Lateral roots contain large amounts of aerenchyma, allowing oxygen diffusion into the rhizosphere (Becker and Asch 2005;Wu et al. 2014); (2) Partitioning of Fe 2+ into organs and subcellular tissues (Moore et al. 2014) of different plants so the plants are more tolerant to the iron excess conditions.
Genomic mapping technologies, such as cytogenetics, molecular genetics, and physical mapping to complete rice genome sequence are essential breakthroughs for uncovering the functional part of the rice genome (rice functional genomic) for many critical complex characters such as tolerance to abiotic stress (Tyagi et al. 2004). Single nucleotide polymorphism (SNP) markers are mainly developed based on next-generation sequencing technology. The fast development of SNP markers through genotypingby-sequencing (GBS) has paved the road to facilitating genomics-assisted breeding through quantitative trait loci (QTLs) and genome-wide association analysis (GWAS) in diverse crops (Basu et al. 2018). GWAS typically focuses on associations between SNPs and dominant traits. Moreover, GWAS is often utilized when we are interested in finding out all the genomic regions that may control a specific role. Association analysis based on linkage disequilibrium (LD) is an efficient way to dissect complex traits and to identify gene functions in rice (Zhang et al. 2016).
Some results of the previous mapping studies have shown that there are several genes or QTLs related to iron toxicity tolerant character. The genes or QTLs spread across multiple chromosomes of the rice genome, including on chromosome 1 that was detected on a physical map position of 25-30 Mb and in chromosome 3 at location of 0-5 Mb (Dufey et al. 2009;Wu et al. 2014). The case in this research is tolerance to iron toxicity. SNP markers that have been confirmed associated with the target character can be used as a tool for assisting in the selection process of molecular markers-based breeding for designing iron tolerant rice varieties. The study aimed to analyze candidate loci related to iron toxicity tolerance in rice by the GWAS approach using custom-designed 384-SNP markers.

Genetic Materials
The genetic materials used were two subsets of different populations of rice. The first was 192 breeding lines (BL) subset population. The progeny lines used in the study came from diverse parents that were used for the crossing of iron toxicity breeding lines. The second was 50 rice landraces germplasm (LG) subset population. List of the two subset populations was presented in Appendix 1.

Designed Custom 384 SNP-Chips
Designed custom 384 SNP-chips were based on the genetic map of several genes associated with character of tolerance to Fe toxicity. A previous study has identified many SNPs (Utami and Hanarida 2014). The SNP primers designed were attached to BeadChip in a 2-micron bead that can hybridize with DNA samples at the PCR annealing time.

Field Assay for Iron Toxicity Evaluation
Phenotype characterization of rice landraces to iron toxicity tolerance was done in the acid soil of upland field in Taman Bogo Experimental Field, East Lampung. The geographic allocation of this field is 50 02" South Latitude and 1050 50" East Longitude, with an altitude of 300 masl. Taman Bogo rice field is a plain to rather plain landform (dominant slope of 0-3%). The soil properties at the experimental sites are shown in Table 1. A field experiment was conducted under swampy inland with low soil pH and also suffered from iron toxicity (Fe soil of 2030 ppm). The soil texture consisted of 29% clay, 33% silt, and 39% sand. The soil macronutrients indicated in a low content of N total (%), Ca, and Mg.
Each rice genotype was planted on two rows of 2 m each, in each plot, with two replications. Tillage was done as a local recommendation by giving NPK 300 kg ha -1 and urea 100 kg ha -1 , at 4 and 7 weeks after planting. The performance of Fe toxicity stress was observed on bronzing assessment, which was scored at 1 month after planting. Mahsuri variety was used as a tolerant control and IR64 as a susceptible control.

Genomic DNA Preparation
Total DNA preparation followed the protocol recommended by Illumina, covering the extraction and purification of DNA from leaves of rice plants. Some 20-50 mg samples of fresh leaves were put into 2 ml microtube which already contains two pieces of stainless steel or tungsten carbide bead of 3 mm diameter and placed in Tissue Lyser Adapter Set 2 x 24. A total of 500 ml of lysis buffer (Thermo Kit) containing 0.25 mg ml -1 RNase was then added in the mixture. Samples were centrifuged 1500xg for 30 seconds and incubated at 56° C for 30 minutes. The samples were then centrifuged back at 6000xg for 20 minutes to separate the DNA from the debris and other contaminants. Purification of total DNA was performed by using Thermo King Fisher Scientific Flex (Thermo Scientific 2011). DNA concentration was standardized by dilution to 50 ng ml -1 as a final concentration.

GoldenGate Genotyping Assay
GoldenGate Genotyping Assay is divided into two main stages, namely the pre-amplification and postamplification stages. Pre-amplification includes activation of biotinylated labeled DNA to prepare the DNA samples for the next post-amplification step. This process included on extension and ligation by the PCR process using the two primers labeled with a fluorescent dye (Primary 1 and Primary 2) and one biotinylated primer (Primary 3), where the Primary 3 allows for marking the PCR products and elute DNA thread containing a fluorescent signal. Post-amplification was finalized by visualizing the BeadChip-signal on the Iscan system. Data visualization was then analyzed to determine the genotype of SNP using Illumina's BeadStudio Gene Expression Module (Illumina 2009).

Data and Association Analyses
Phenotypic data were analyzed using two way ANOVA to test the effect of Fe toxicity stress and genotype factors. The data were analyzed using the XLSTAT 19.5 software program (www.xlstat.com). Association analysis was done on the whole population of 242 individual genotypes and subpopulation separately, consisting of 192 genotypes of Fe tolerant breeding lines and elite varieties and 50 accessions of rice landrace. Association between SNP markers and phenotypic data was tested using the General Linear Model (GLM) in the TASSEL (Trait Analysis by aSSociation, Evolution and Linkage) v. 3.0 software program (Bradbury et al. 2007). However, researchers must contend with the confounding effects of both population and family structure.

Phenotypic and Genotypic Diversity for Association Analysis
The BL subset samples (A) have a different performance to LG subset samples. The BL was homogenized in each line, although they were diverse in agronomical characters, such as plant height. The overall responses of both populations varied on Fe toxicity tolerance based on bronzing score. This variation can be seen in each set of different populations. This is due to the influence of diverse genetic backgrounds (Figure 1). The phenotypic performances of 50 accessions of LG population showed abundance on morphological variation. It was contrasted to 192 BL population, although they were developed from a broad genetic background ( Figure 1A). It showed that the genetic diversity of BL had been reduced compared to LG.
The association analysis between phenotypic and genotypic data in the BL population detected three LD blocks spreading over in several SNPs markers locations. It was in contrast with the LG population which showed only one LD block ( Figure 1B). These results were relevant to the LD map that showed three blocks and one block for BL and LG populations, respectively. LD block map has measured the strength of the correlation between markers caused by their shared genetic history (Bush and Moore 2012). Due to the BL population have the same genetic history on Fe toxicity tolerance breeding program, they have a lot of pairs of SNP that correlated with an allele of another SNP and associated with the Fe toxicity tolerance alleles. The different conditions were showed on LG population; they have breeding naturally as landraces originated from the swampyland, which only had one unsaturated LD block. These results are the critical thing in identifying the candidate loci associated with markers (Soto-Cerda and Cloutier 2012).

Analysis of Candidate Loci Associated with Iron Toxicity
Identification of some SNPs markers included in the set of 384 SNP-chip-2014 significantly related to the Fe tolerant character based on the field testing for two seasons in two different locations, namely Karang Agung, South Sumatra (2014) and Taman Bogo, Lampung (2015), which showed that some selected SNPs markers spread across on the 12 chromosomes of rice genomes (Table 1). These closely flanking markers based on the genetic position contained in several gene loci have been identified in earlier research. The association analysis results indicated there were several SNPs as candidate loci associated with Fe toxicity tolerant characters. The genetic position of the SNP markers is consistently significant on the set of the population and with different mapping association methods, i.e. Generalized Linear Models (GLM) and Linear Mixed Model (MLM).
Furthermore, the tracking analysis of gene functions following the genetic position of significant SNP markers at Rice Annotation Project (RAP) and The Institute for Genomic Research (TIGR) databases showed that the candidate loci were located on chromosomes 1, 2, 3, 4 and 7 of rice.

A B
the plant is more tolerant to Fe 2+ excess conditions. IRT gene expression occurs in the leaves and stems (Ishimaru 2006).
The phenotypic performance of some test plants such as IR54, a tolerant variety, showed a bronzing score of 2-3, whereas IR64, a sensitive type, had a 9-bronzing score. IR54, a tolerant variety, had an excellent agronomic performance both during the vegetative or generative stages. It is different from IR64 that during the vegetative stage, high growth is hampered (Abu et al. 1989). IR64 also has a problem in tillering development (Cheema et al 1990). The effect of Fe toxicity was also seen when the plants entering the end of the vegetative stage or at initial of the generative phase. Fe toxicity inhibited the panicles formation and even the number of grains in each panicle (Singh et al. 1992). Fe toxicity also causes the plants to be sterile or disrupts the flowering (Virmani 1977).
In the high Fe conditions, root performance of IR54 and IR64 were not significantly different (Figure 2). This indicates that both varieties have Fe 2+ transport activity by the same gene, IRT. However, IR54 can partition the Fe 2+ absorbed into the tissue that is not done by IR64. Therefore, based on the candidate loci analysis, allele groups in LD blocks on chromosome 1 are thought to play a role in the Fe 2+ partition activity as a part of Fe tolerance mechanisms in rice.

Chromosome 2 and 3
Based on association mapping analysis, chromosome 2 and 3 had two groups of significant alleles to bronzing levels of landrace samples tested ( Figure 3A). The alleles group on chromosome 2 were mapped in position of 26.3 kb, in accordance with the genetic map of QFE-TOX-2-1 (Shimizu 2009) and 31.8 kb, in accordance with qFETOX-2 which mapped in the RIL (F8) population of IR29 (sensitive) and Pokkali (tolerant) (Wu et al. 2014 ). Observation of the cross section of the root showed that Pokkali had Table 2. SNP markers selected from 384 SNPs-chip-2014 associated with iron tolerance character (bronzing score) in rice.

SNP ID Chromosome
Genetic map P-value Reference TBGI067836 1 40,320,704 0.01384305 Wu et al. (1997Wu et al. ( , 1998 Figure 2. A. The three candidate loci on LD block accordance with the genetic map position of the significant SNP markers on the total 38 SNP markers spread in chromosome 1. B. Plant performance of IR54 (tolerant rice variety) and IR64 (sensitive rice variety) at two months after planting in high Fe conditions. C. Root performance of IR54 and IR64.
A aerenchymal tissue higher than IR29, both in the initial conditions or under Fe 2+ stress conditions ( Figure 3B). Aerenchyma is the parenchymal tissue that holds the air with the structure of an ample space between cells. These plant tissues contribute to the internal oxidation process in the plant (Colmer 2002 ). Great parenchyma will increase the oxidation of the roots, which will also further enhance the ability of Fe 2+ exclusion on the roots of tolerant plant Pokkali (Wu et al. 2014). Thus, the plant can limit the absorption of Fe 2+ . Some rice landrace samples tested in the field ( Figure  6C) showed the correlation between root and plant performance with a bronzing score, which is one of the Fe toxicity parameters. Tolerant plants (bronzing score 1-3) have longer and more robust roots than those of sensitive plants (bronzing score 9). This indicates the existence of Fe 2+ prevention mechanisms into the roots to avoided damage to other plant parts.
As it is the case on chromosome 2, LD blocks on chromosome 3 were also detected in two genetic positions, on 3.2 kb (Wan et al. 2003;Dufrey et al. 2012) and 10.9 kb (Inoue et al. 2003). Based on the analysis of the candidate loci, allele groups in LD blocks on chromosome 2 and 3 presumed play a role in the Fe 2+ exclusion activity, parts of tolerance mechanisms on inland contained high Fe.
Some of the genes included in the LD blocks in chromosomes 4 and 7 have been known to play a role in the partitioning and exclusion of Fe 2+ in Fe tolerance mechanism ( Figure 3B) (Tsai and Schmidt 2017). The role of association results in this study was implicating to candidate loci detected, which contributed to Fe toxicity tolerance as shown in Table 3. The performance of the test plants was observed in the field on two varieties, i.e. IR64 (sensitive control) and Mahsuri (tolerant control) ( Figure  3C). Mahsuri has strong roots and good shoot in Fe toxicity conditions. These varieties are expected to have tolerance mechanisms at the root level, i.e. the regulation of Fe 2+ absorption and in the shoot level on Fe 2+ partition capabilities. This is as proposed by Saikia and Baruah (2019), who reported that Mahsuri on 350 ppm of Fe 2+ could control Fe 2+ absorption and increase superoxide dismutase (SOD) accumulation.

CONCLUSION
Iron tolerance response of 242 rice accessions based on phenotypic experiment in Taman Bogo, Lampung varied, both on BL and LG populations. Association mapping on the BL population detected three positions of LD blocks around significant SNPs. Based on the candidate loci analysis, there were three loci identified. The first LD block was in chromosome 1, mapped on the AtIRT gene locus, as well as QTL of qFETOX1 and qFETOX 1-3. These loci assumed to play a role in Fe toxicity tolerance through the Fe partitioning mechanism. The second LD block located in chromosome 2, mapped on the qFETOX-2-1 and qFETOX-2 loci, as well as in chromosome 3, on qFETOX-3 locus and OsNAS1 and OsNAS2 genes. These allele groups were predicted to contribute to the Fe exclusion mechanism.The third LD block is in chromosome 4, mapped on OsFRO2 gene and qFETOX-4 and chromosome 7 on OsIRT2 and NAS3 genes. These allele groups presumably contribute on both partitioning and exclusion of Fe. GWAS approach can probably detect allele groups that contribute to the  Table 3. The predicted role of candidate loci identified in this study referred to previous reports based on the association analysis.

LD block
Chromosome Candidate loci Proposed mechanism Description of association I 1 AtIRT qFETOX1 qFETOX1-3 Partitioning of Fe 2+ Fe tolerance mechanism was contributed by partitioning activity and no absorption exclusion at the root level. The rice lines contained LD block were probably more adapted on the tidal swampland area because the toxic Fe 2+ will decrease through the oxidation process.
II 2 qFE-TOX2-1 Exclusion of Fe 2+ Fe tolerance mechanism was contributed by exclusion of Fe 2+ activity in root and no partitioning in tissue level.
The rice lines contained LD block II were probably more adapted in a lowland swamp area Fe tolerance mechanism was contributed to both Fe 2+ partitioning and exclusion in root level activities.
The rice lines contained LD block III will have broad tolerance in diverse swampland conditions. qFETOX4 7 OsIRT2

OsNAS3
Fe tolerance mechanism, although the function of those candidate genes/QTLs should be verified. The study implies a valuable result for determining the strategy for developing Fe-toxicity tolerance rice for specific swamp land type through breeding programs.

ACKNOWLEDGEMENT
The research was funded by the Indonesian Center for Agricultural Biotechnology and Genetic Resources Research and Development. Thanks to the research team of the Indonesian Center for Rice Research (ICRR) and Taman Bogo Experimental Station, Lampung, for teamwork and collaboration.

AUTHORS CONTRIBUTIONS
DWU was the main contributor, responsible for designing the research, analyzing-interpretating data, and writing the manuscript. IR, LC, and SN assisted in data analysis. Sb and S supported genetic materials and conducted a field experiment.