Rice is a major staple crop in the world and demand for rice is increasing every year (Ray et al., 2013). Biotic stresses continue to be the constraint in rice production and becoming severe in the climate change regime. One of the most identifiable major biotic stresses is the blast disease caused by Magnaporthe oryzae. It infects rice leaves, nodes, color, and panicles at different stages of crop growth thus, decreasing the overall yield. In India, neck blast is becoming severe in many agro-ecological zones and causing more threats to the rice production (Laha, 2017). Although 100 major blast resistance genes (R-genes) have been identified, mapped and their tightly linked DNA markers are available (Miah et al., 2013), only one major gene (Pb1) has been reported for neck blast (Hayashi et al., 2010). In addition to this, rapid evolution and high variability of Magnaporthe with the rise in temperatures relates to its high pathogenicity. Hence identification of durable resistance sources and resistance genes is a continuous process in rice improvement programs.

Landraces are genetically dynamic and display equilibrium with both the environment and pathogens (Harlan, 1975). Many potential landraces of rice are being replaced by high yielding varieties to meet the food requirements. Despite being less productive they are known to have a high genetic variance for several biotic stresses (Hanamaratti et al., 2008), so they can be explored for rice improvement. North-East (NE) India, known to have highly diverse rice germplasm and the landraces belonging to this region are expected to have high genetic variability for many biotic stresses including the rice blast. Many rice landraces of NE India were not yet characterized for blast resistance.

Genetic diversity among the populations and their genetic relationships aids in conservation and parental selection in the improvement programs. Identification of populations with a high level of genetic variation will become a valuable resource for broadening the genetic base since it enables the identification of superior alleles for several traits including blast resistance. Association mapping is the best approach to identify the chromosomes segments harboring genes/QTLs (quantitative trait loci) controlling key traits and the genotype–phenotype associations in germplasm. For the successful application of association mapping, understanding of population structure is essential to avoid type I and type II errors between molecular markers and traits of interest (Zhang P. et al., 2011). Determination of population structure also helps in categorizing populations into subsections which are genetically related. The rapid advances in marker technology along with morphological traits can precisely estimate the genetic diversity among the populations (Kumbhar et al., 2015). For estimation of genetic diversity, population structure, linkage disequilibrium, and mapping various traits, molecular markers have been successfully used by many researchers (Ried et al., 2011; Nachimuthu et al., 2015; Anandan et al., 2016). SSRs aid in accurately estimating the genetic diversity among germplasm and are reported to be more efficient than single-nucleotide polymorphic markers (SNPs; Das et al., 2013; Singh et al., 2013; Nachimuthu et al., 2015). Development of a core set representing the total variation of the population with novel and superior traits is essential for successful utilization of the population. Core and mini core sets of rice were earlier made using morphological traits and molecular markers (Zhang H. et al., 2011; Liu et al., 2015). However, no core set is available exclusively for the NE landraces. Evaluation of genetic diversity among NE Indian germplasm was reported earlier (Das et al., 2013; Roy et al., 2014, 2015) but very few studies exist on the understanding the diversity in relation to the blast resistance (Ghaley et al., 2012; Mahender et al., 2012).

In the current study, we have systematically performed phenotyping for blast resistance in a span of two years. Phenotyping along with gene profiling for major blast genes identified novel genetic resources which might contain the novel resistance genes. For clear understanding, the genetic diversity of NE landraces were analyzed with distance as well as model based diversity approaches. To apply the existing diversity in breeding programs, a core set was derived. Using 232 landraces, the marker–trait association was identified for blast resistance as well as yield-related traits. The information generated in the present study has a lot of value in the current rice improvement programs.

A total of 232 landraces (Supplementary Table S1) were collected from the NE region of India (Figure 1) for the current study. Along with the landraces, three varieties of Basmati, five varieties of Indica, seven varieties of Temperate Japonica, and five varieties of Tropical Japonica were also selected for the comparison of landrace alleles. The line harboring particular blast resistance gene was considered as a positive control, and the susceptible cultivars namely BPT, Swarna, and CO39 which do not have any blast resistant gene were considered as negative controls for gene profiling.

Successful rice breeding programs require the presence of genetic diversity in the existing germplasm. The current study focuses on ascertaining this phenomenon in NE Indian landraces. We used a completely unique and unexplored germplasm for identifying the best donors and markers associated with blast resistance. We demarcate the diversity into a manageable core set.

Systematic screening for blast disease at two different locations for two consecutive seasons led to the identification of 91 landraces resistant to both leaf and neck blast. Blast screening at different locations increases the consistency of phenotyping as different virulent blast isolates have diverse pathogenic behavior (Madan et al., 2012). Previously a similar strategy was used for identification of best introgression lines for blast resistance (Devi et al., 2015). A mixture of blast isolates from geographically distinct locations (Hyderabad and Manipur) led to the identification of stable QTLs in one of the NE landrace (Aglawe et al., 2017). Neck blast is becoming more problematic than the leaf blast at several locations in India (Aglawe et al., 2017; Laha, 2017) so, it has become imperative in breeding programs to choose the accession having resistance for two phases of blast resistance (Abhijeet et al., 2013). So far only one neck blast resistance gene (Pb1) was identified (Hayashi et al., 2010) and many of leaf blast resistant accessions may not hold the resistance for the neck blast, hence screening for neck blast was given a priority. Moreover, the NE hilly areas have compatible climatic conditions for a natural infestation of neck blast.

In addition to the phenotyping, gene profiling using SNP, SSR, STS, and indel markers was also performed so as to identify the presence of valuable genes to the blast resistance (Cho et al., 2008) (Supplementary Table S2). These results indicated that Pi9 was major gene followed by Piz and Pizt existing in landraces. The Pi9 is predicted to be persistent in landraces from a long time as it was originated from the wild species (Oryza minuta) of rice. Anupam et al. (2017) identified the frequency of seven blast genes ranging from 0 to 80% in 74 local landraces collected from Tripura by profiling with gene-based STS and SNP markers. Prevalence of Piz among 32 NE Indian landraces by gene profiling with STS and SNP markers was reported (Imam et al., 2014). Shikari et al. (2013) also evaluated the 54 landraces collected from different parts of India for the presence of Pita and Pita2 genes with two gene-based and three linked markers and found 11 landraces containing Pita and nine landraces containing Pita2 genes. Gene-specific markers of eight blast resistance genes (Pib, Piz, Pizt, Pi9, Pi40, Pi5, Pia, and Pita) were also explored by Mahender et al. (2012) for characterizing the various accessions of Manipur (another state of NE India) and found landraces containing two to seven blast resistance genes, which again indicates the richness of blast diversity in landraces of NE India. In the current study, the landraces used were entirely different from the previous work (Mahender et al., 2012; Imam et al., 2014; Roy et al., 2015), we also checked their presence in the national gene bank (National Bureau of Plant Genetic Resources, NBPGR, New Delhi) and found 189 landraces that does not exist in it. Around 85% of landraces (197) harbors blast genes ranging from 1 to 7, indicating the rich diversity for blast resistance and these landraces serves as good genetic resources for blast resistant genes. Interestingly, seven resistant landraces did not show the presence of any of the tested genes, may contain the novel genes (Supplementary Table S1 and Figures 2, 3). Identification of new genes having durability and its transfer into a cultivable variety may increase its resistance toward the disease since most of the identified genes are losing their effectiveness to blast disease (Das et al., 2012).

Genetic diversity analysis was done for determining the diverse nature of the landraces. To precisely differentiate closely related individuals, 120 SSR markers which are well distributed in the genome were used in the current study. The extent of polymorphism was detected by calculating PIC values and for evaluation of genetic diversity across the chromosomes; an allelic number was also calculated. Average PIC value of markers was 0.44 with an average of three alleles per locus (Supplementary Table S3). An average PIC value of 0.47 was reported in 74 Tripura rice landraces by using 30 SSR markers (Anupam et al., 2017). Roy et al. (2016) found a PIC value of 0.62 with the NE Himalayan landraces. Yang et al. (1994) reported average PIC value of 0.58 in 238 landraces collected from India, China, Japan, and IRRI. Giarrocco et al. (2007) also found average PIC value of 0.69 in 68 accessions of Argentina. A significant amount of rare alleles were found (15%) which might contribute to the overall genetic diversity of the population. Nachimuthu et al. (2015) reported the presence of 5% rare alleles in 192 rice germplasm upon screening with 61 SSR loci. Anupam et al. (2017) reported the existence of 11 rare alleles in 74 germplasm lines upon analysis with 30 SSR loci.

In the present study, a gene diversity of 0.51 was observed among the total population (232 landraces), which is comparable to most of the diversity panel consisting of global accessions (gene diversity of 0.5–0.7) (Ni et al., 2002; Garris et al., 2005; Ali et al., 2011; Roy et al., 2015, 2016). Gene diversity of NE landraces is higher than overall gene diversity of rice core collection (0.544) from China, Philippines (Zhang H. et al., 2011) and US accession panel (average gene diversity of 0.43; Agrama and Eizenga, 2008). These results clearly indicating that the NE landraces of India are showing the high genetic diversity in rice that exists in India (Supplementary Table S3).

The distance based clustering analysis grouped the landraces into two major clusters (Figure 4). Choudhury et al. (2013) used the distance based clustering method to group the 100 cultivable rice varieties released in different states into their respective decade periods. Similarly, rice genotypes collected from Tamil Nadu were classified into indica and japonica types based on the distance based clustering (Vanniarajan et al., 2012). In addition to the distance based clustering of the population, the genetic architecture of diverse material was also done by the model based structure analysis, as it offers better architecture of population with molecular markers (Powell et al., 1996; Horst and Wenzel, 2007; Varshney et al., 2007). Several researchers have analyzed the population structure by implementing model-based approach using Structure for numerous rice accessions (Garris et al., 2005; Agrama et al., 2007; Zhang et al., 2007, Zhang P. et al., 2011; Agrama and Eizenga, 2008; Jin et al., 2010; Ali et al., 2011; Chakhonkaen et al., 2012; Courtois et al., 2012). The present study classifies the landraces into two subgroups SG1 and SG2 using the threshold value >80%. Similar results were observed in grouping 3024 Chinese rice landraces (Zhang et al., 2009), 64 landraces of NE Himalayan region of India (Roy et al., 2016), and 192 rice accessions of varied origin (Nachimuthu et al., 2015). But, several studies were reported on the grouping of rice accessions which varied from two to eight groups (Garris et al., 2005; Agrama et al., 2007; Zhang et al., 2007, 2009; Zhang P. et al., 2011; Ali et al., 2011; Chakhonkaen et al., 2012). Roy et al. (2015) divided 107 NE landraces into three different subgroups by structure analysis using 40 SSRs. Das et al. (2013) used 23 SSR markers and distinguished 91 rice lines (which include 83 NE landraces and eight checks) into four subpopulations. Thirty-seven chakhao landraces of NE were also divided into six subgroups based on genetic structure analysis (Roy et al., 2014). The current study grouped 232 accessions based on the maximum membership probability of >80%, into 33 accessions under SG1, 158 accessions under SG2 and 41 accessions as an admixture (Figure 5). Further population structure analysis in SG2 classified the landraces into four subgroups indicating that independent structure analysis defines the precise genetic structure of the accessions. The results of the unweighted neighbor-joining tree constructed on the basis of UPGMA and PCoA harmonized with the structure analysis of the landraces. The PCoA cited the resistant and susceptible genotypes into four different quadrants at different positions.

AMOVA is useful for studying molecular variance within the species. It is used to study the patterns and degree of relatedness revealed by multidimensional scaling and the clustering dendrogram. It also summarizes the population structure, while remaining flexible enough to accommodate different types of assumptions about the evolution of the genetic system. AMOVA of the population showed greater variability among the individuals within the populations. The variation observed among the individuals within populations was higher (75.1%) than the variation present between the populations (24.9%). The F_ST value of 0.27 indicated that the higher divergence was observed between subgroups. Nachimuthu et al. (2015) upon analyzing 192 rice accessions (landraces and varieties collected from different parts of the world) and grouped them in two subgroups and reported 14% variations among the groups and 86% within the group. Similarly, Zhang et al. (2009) also grouped the 3024 rice accessions, into two subgroups and found the variation among the groups which was little higher (36.65%) and variation among individuals within populations was 40.80%.

The development of core set is very much essential for better management of germplasm lines in crop improvement strategies. It is helpful to precisely characterize, explore and conserve resources, monitor the genetic drift during preservation and identify gaps in genetic diversity (Frankel and Brown, 1984). Choudhury et al. (2014) developed a core set of 701 accessions (10%) among 6984 germplasm lines collected from NE India using molecular markers and agronomical traits. Zhang H. et al. (2011) established a core set of 189 accessions from 4310 Chinese accessions which include landraces. In the present study, a core set of 14% (33 landraces) was developed which represents the existing entire genetic diversity of the total landrace collection (Supplementary Table S4A and Figure 6). The core set developed indicates the homogeneous distribution ranges of the phenotypic traits as MD% is less than 20% and CR% is larger than 80% (Kim et al., 2007). The VD% and VR% present within the range represents the diversity as the original collection (Hu et al., 2000). This shows that the core collection represents the entire collection. At present, there is no universally accepted method for the development of core set as various factors such as sampling percentage, data type, number of traits observed, genetic diversity of germplasm, grouping method, and sampling method will play in demarcating the core group (Upadhyaya et al., 2006; Diez et al., 2012; Rao et al., 2012).

Association mapping is a novel approach which helps in identifying marker–trait relation based on linkage disequilibrium (Zondervan and Cardon, 2004). Based on GLM, we identified 20 marker associations for blast disease and other agronomic traits. Of which, eight markers loci are novel since, in their location, QTLs for those traits were not reported. These newly identified trait-associated SSR markers can be used for the discovery of novel genes/QTLs. While with MLM analysis, only 17 marker–trait associations were found and all these associations were reported in the GLM. Associations detected in GLM were not found in MLM, since in later analysis effect of kinship may play a role. The markers RM5499 and RM190 which are previously reported for grain yield also showed association in the present study and in the same way two markers (RM286 and RM6051) for plant height. Few markers showed association with more than one trait which is of great interest in breeding programs. Two markers RM21936 and RM3787 were newly identified to be associated significantly (P < 0.05) with leaf and neck blast disease (Table 2). These markers may be helpful for mapping genes for neck blast with durable resistance which does not exist in the currently available germplasm. Exploring novel gene pool for neck blast resistance and identification of novel genes which can offer resistance to leaf and neck blast resistance has become quite essential. Anandan et al. (2016) reported 16 marker associations in 629 rice genotypes using 39 SSR markers for identification of genes associated with early seedling vigor using GLM. In an association analysis of yield and grain quality traits in Embrapa Rice Core Collection (ERiCC) of 242 accessions (Inbred lines, cultivars, upland, and lowland accessions) from Brazil with 86 SSR markers and field data, eight markers were identified which were associated with four different traits (Borba et al., 2010). Twenty marker associations were identified for 29-grain metabolites in rice using 218 SSR markers in a set of 48 rice germplasm lines from the Chinese core collection (Lou et al., 2011). Twenty-six SSR markers associated with blast resistance (P < 0.01) in a set of 276 indica landraces from China and few from different parts of the world were reported (Wu et al., 2016). The identified markers associated with specific traits can be validated for their effectiveness in various genetic backgrounds and those can be helpful to the breeders in the pyramiding of QTLs from different sources through marker-assisted selection. Hence association mapping became an unleashing tool for identification of markers linked to different traits.

Host and pathogen arms race is a continuous phenomenon in evolution. The premise of collecting, analyzing and characterizing unexplored NE landraces is for identification of donors containing multiple blast resistance genes which can be readily used in the breeding programs. The unique seven landraces (Meghalaya lakang, Chingphourel, Manuikhamei, Kemenya kepeyu, Wainem, Thekrulha, and Koyajang) having resistance to leaf and neck blast can be explored for the discovery of novel genes for broadening the gene pool to combat the pathogen competition. Distance and model based diversity analysis are highly correlated and genetic architecture of diverse material was effectively determined. The rich diversity existing in the landraces were divided into manageable subgroups and identified the molecular variance present among the populations and individuals which can be harnessed in the selection of distant individuals. To utilize the diverse landrace germplasm, an efficient core set which represents the entire diversity was made. Further, association analysis using a core set identified novel eight markers associated with the blast and other agronomically important traits, these markers are the candidates for the validation in the larger populations. The information emanating this work will help in conservation and utilization of valuable rice germplasm of NE India.

MSM designed all experiments, developed the structure and arguments for the paper, made critical revisions, and approved the final version. BU and BV carried out the major work and prepared manuscript and contributed to the refinement of the manuscript. PSK helped in analyzing the data and SRD helped in data collection of agro-morphological traits. PS reviewed and approved the final manuscript. VPB, SK, SKS, and PKS helped in the collection and phenotypic screening of rice landraces for neck blast resistance at the North-East region. MSP helped in the screening of leaf blast resistance at IIRR. All authors have read and approved the final manuscript.

The authors declare that the research was conducted by the support of DBT, Government of India.