Genetic analysis of yield and yield contributing traits in rice (Oryza sativa L.) BC2F3 population derived from MR264 × PS2

Abstract High yield potential in rice is indirectly determined by yield-related traits. These traits are complex and regulated by several genes whose expression is affected by environmental conditions. It is of great importance to disclose the genetic relationships between yield and its yield components for multi-trait improvement in rice. Therefore, the present study aimed to investigate the genetic variability and inheritance patterns of yield and yield attributed traits in BC2F3 rice lines to identify the ideal lines from the selection. A total of 36 improved versions of blast resistant plants in the BC2F3 population used in this study were developed from a single cross between a high yielding mutant rice variety but susceptible to blast, MR264, and Malaysian local variety donor for Pi-7(t) and Pikh blast resistant genes. Analysis of variance showed that all traits were significantly different for lines except grain length and grain width. High heritability and genetic advance were recorded for plant height, number of tillers, filled grain, 1000-grain weight and seed setting rate. A significant and positive correlation was recorded with most evaluated traits, except for grain length and grain width. Thirty-six BC2F3 lines were clustered into four major groups and the first three principal components (PC3) contributed 71.13% of the total variation, with 1000-seed weight, yield/hill and filled grain being the main discriminatory characters. There was an adequate genetic variability in the lines, and 1000-grain weight, yield/hill and filled grain traits could be considered for indirect selection in breeding programs in next generations.


Introduction
Rice is a primary source of nourishment that contains carbohydrates, numerous vitamins and minerals. It contributes over 30% of the caloric intake per capita in the world. Along with surging world population, to fulfill the population demand, rice production must grow by 50% as it is prone to climate instability [1]. In Malaysia, rice production is hindered by blast disease caused by a pathogenic fungus, Magnaphorthe oryzae, which has resulted in more than 50% of yield losses [2][3][4]. Development of blast resistant rice genotypes with high yield is therefore essential to aid resource-poor farmers who rely on rice as their supply of energy, to keep up with the surge in demand of the growing and increasingly affluent human population and ensuring global food security.
Increasing rice production can be fully exploited with various gene pool resources and its suitable environment to obtain desirable high yields and some stress tolerance. The development of new rice lines with the improvement of the yield of required information related to the combination of parental abilities and gene action involved in the expression of maturity and morphological characteristics of plants [5]. The selection of assorted parents for hybridization must therefore be based on the ability of the different parental lines to combine. In preparing the program efficiently and effectively, the analysis of genetic variance and the mode of inheritance of quantitative and qualitative traits are of primary importance. Genetic properties of the breeding material and the environmental conditions under which experiments are conducted determine the heritability of any traits [6]. Thus, a greater range of genetic diversity, high heritability and high genetic advance play a key role to enhance rice yield.
Morphological traits have been used to assess the genetic variation and relationships among populations of rice [7,8]. It is considered as the primary step in the classification and evaluation of germplasm [9]. These qualitative characters are crucial for plant description and influenced by consumer preference, socio-economic scenario and natural selection. Several morphological characters are the major determining factors of rice grain yield including number of panicles per hill, the number of filled grains per panicle, and weight of filled grain per hill [10,11].
However, morphological traits have a few drawbacks including low polymorphism, low heritability, low expression and vulnerability to environmental influence [12,13]. Diversity and interrelationships among yield parameters are the key component in rice breeding strategies. It is therefore imperative to examine the level of genetic transmissibility and gain of yield contributing characteristics to forecast selection response for further improvement [14]. In this scenario, a heritability estimate is required since it demonstrates the genetic participation in character transmission from one generation to the next [15]. It is critical to understand the relationship between yield and various yield attributing qualities in order to improve the yield attributing traits utilising various breeding approaches. This knowledge will assist the breeder in developing a good yield improvement plan.
In addition, correlation studies are also important to provide a better understanding of the relationship between grain yield and other yield attributes [16]. Cluster analysis which groups genotypes based on their similar performance in terms of many variables, is becoming increasingly popular as a tool for diversity study [17]. Preliminary works have identified a total of 36 selected homozygous blast resistant lines (carrying Pikh and Pi7(t)) from a cross between MR264 × Pongsu Seribu 2 [18]. Thus, the present study was designed to determine the genetic diversity, the character inheritance pattern of grain yield in BC 2 F 3 population and to identify the best lines from the selection for further breeding new blast resistant rice varieties.

Plant materials and experimental design
An experiment to assess the genetic diversity of yield related characters among the BC 2 F 3 population was carried out from September 2017 to April 2018 in the greenhouse at Malaysian Nuclear Agency greenhouse, Kajang, Selangor, Malaysia. The study site is located between 2°91′N and 101°78′e at an altitude of 31 m above sea level. The materials in this research were 36 homozygous lines of BC 2 F 3 generation seeds from the pedigree selection resulting from a single cross between local varieties, namely MR264, a high-yielding and semi-dwarf rice variety but blast susceptible, and Pongsu Seribu 2, a well-known traditional Malaysian variety that shows a wide spectrum of resistance to blast fungal isolates [3]. The breeding scheme for developing the population is summarized in Figure 1. The experiments were performed in Complete Randomized Design (CRD) with five replications by using MR264 as check varieties. Seedlings were transplanted in individual pots and each plant was grown in a standard plastic pot (28 cm × 25 cm) with five replications. NPK fertilizer (150 kg/ha Urea, 100 kg/ha SP36 and 100 kg/ha KCl) was applied twice, with the first fertilization at 10 days after planting (DAP) and the second fertilization, at 30 DAP. Water management and control of weeds, pests and diseases was intensively done for rice cultivation as recommended by the IRRI International Rice Research Institute [19].

Assessment of agronomic performance
Agronomic parameters were assessed using selected lines carrying blast-resistant alleles. The following quantitative traits were considered for data collection: days to maturity (DM, days), plant height (PH, cm), number of tillers per hill (NoT, n), panicle length (Pl, cm), number of filled grains (FG, n), grain length (Gl, mm), grain width (GW, cm), yield per hill (Y, g), 1000-grain weight (100 GW, g) and seed setting rate (SS, %). A standard evaluation system (SeS) recommended by the IRRI International Rice Research Institute [19] was followed to record the data.

Statistical analyses
The analysis of variance and comparison of mean differences among treatment was performed using SAS version 9.2 at probability level of 0.05. The phenotypic and genetic parameters including the genotypic and phenotypic variance, genotypic and phenotypic coefficient of variance, heritability (broad sense), and the expected genetic advance (GA), were estimated using the equations of [20] and [21]. The data were calculated using the following formula: where GMS is the genotypic mean square, EMS is the error mean square and r is the replication number.
where X is the population mean.
where h 2 b is heritability in broad sense; K is the selection differential, which is 2.06 at 5% selection intensity; σ p is the phenotypic standard deviation.
Association analysis among the 10 quantitative traits of 36 rice lines was performed using Pearson's correlation coefficient. Statistical analysis system software (SAS version 9.1, SAS Institute, 2001) was used to analyze the phenotypic data.

Cluster and principal component analysis (PCA)
Genetic diversity such as cluster analysis and PCA was performed using NTSYS-PC software (version 2.1) based on euclidean distance method, Dice's and Jaccard's similarity coefficient following the methods described by oladosu et al. [24]. Unweighted pair group method with arithmetic means (UPGMA) algorithm and SAHN clustering was used to analyze genetic relationships among the rice lines. eIGeN and PRoJ modules of NTSYS-pc and SAS version 9.2 were used to calculate the PCA of 36 rice lines.

Estimation of yield and yield attributed parameter
Among the BC 2 F 3 lines, almost all the agro-morphological traits exhibited highly significant variation (Supplemental Table S1). Traits of significant value consisted of days of maturity, plant height, number of tiller per hill, panicle length, number of filled grain per panicle, yield per hill, 1000 grain weight per hill, and seed setting rate. As mentioned by Roy and Shil [22], the existence of high variability in breeding lines will increase the likelihood of producing desirable recombinants in successive generations. However, the differences were not significant for GW and Gl. Thus, there is possibility of genotype improvement through selection.
P24 exhibited the highest DoM (123 days) while P13 showed the lowest DoM (112 days). The highest PH was recorded in P19 (104 cm) while P4 showed the lowest PH (76 cm). The NoT varied from 10 to 28. The highest NoT (28) was found in P6, which was statistically similar to P3 and P4. The highest Pl was observed in P4 (26 cm) followed by P2 (25.67 cm) while the shortest Pl (21 cm) was found in P3 and P26, which was statistically similar to P20 and P33. A significant difference was recorded for FG among all lines, with values that ranged from 210 to 96. The highest Y was found in genotype P6 (44.67 g), followed by P4 (43 g), whereas the lowest value (33.67 g) was observed in P33 (Supplemental Table S1). The highest 1000-GW belonged to P2, whereas P16 and P33 had the lowest amount of 1000-GW. The highest SS was recorded by P3 (96.33%), which was statistically similar to P5, P11 and P12. P36 showed the lowest SS (54.33%). The Gl and GW also varied among the progenies, ranging from 7.67 to 11.33 mm and 1.83 to 2.07 mm.

Determination of genetic variations
The genetic parameters such as GVC, PCV, heritability and genetic advance are presented in Table 1. The results revealed a broad range of variability among the 36 BC 2 F 3 lines for 10 quantitative traits. The phenotypic variance (σ2P) was greater than the genotypic variance (σ2G) for all the traits. Similarly, the phenotypic coefficient of variation (PCV) was also higher than the genotypic coefficient of variation (GCV). NoT and FG showed the highest PCV and GCV values with 28.72 and 26.24% and 26.19 and 25.60%, respectively. In contrast, the lowest PCV can be found in the characters of DoM (5.57%), GC (8.24%) and Y (9.23%). Similarly, DoM (3.76%), Pl (6.84%), Gl (6.42%), GW (0.88%) and Y (7.99%) were the traits recorded for the lowest GCV. In this study, most of the traits displayed a high heritability, except for grain length and grain width, which exhibited the lowest heritability percentage with 12.23 and 1.15 respectively. In addition, high estimates of heritability were also observed for FG (95.12%), PH (86.40%) and NoT (83.17%). The genetic advance calculated in this study ranged from 51.42% for FG to 0.20% for GW.

Cluster analysis
A UPGMA dendrogram was constructed among 36 selected progenies along MR264 by calculating the euclidean distance. The dendrogram displayed the similarity coefficient range from 2.01 to 22.65 ( Figure 2). In the present study, 36 lines with MR264 were grouped into four clusters based on the 10 quantitative traits at 12.33 dissimilarity coefficient. The cutoff point was set at 12.33. Among the clusters, cluster IV consisted of 17 lines, followed by cluster I with 8 lines and cluster II with 6 lines (Supplemental Table S2). We observed that 7 lines (5, 6, 7, 20, 21, 23, 24) were grouped with the parental variety MR264 as a result of high dissimilarities based on the studied agro-morphological traits. The yield per hill is almost similar across all clusters except for Cluster IV with the lowest yield produced ( Table 2). The number of tillers and panicle length was relatively higher in cluster III but the days of maturity were almost alike across the cluster. The progenies in Cluster IV were found to have the highest values for plant height while Cluster I recorded the lowest plant height value. Clusters II and IV exhibited the lowest filled grain per panicle, with 146.45 and 122.17, respectively. Similarly, the lowest seed setting rate was demonstrated from the progenies grouped in the same Cluster. The grain width ranged from 1.97 to 1.98 mm, with Cluster I and II showing the highest grain width. Cluster IV recorded the highest grain length with 10.38 mm.

Principal component analysis (PCA)
PCA produced a similar result to that demonstrated in the cluster analysis ( Figure 3). This indicated that the data obtained from this experiment were precise, accurate and reliable. The PCA analysis demonstrated that the three principal components explained 71.13% of the total variance and were highly correlated. eigen   vector analysis suggested that 39.18, 19.98 and 11.97% of the total variation of the measured traits could be explained by the first 3 principal components (Table 3). In the first PC: 1000 GW (0.81), Y (0.79), FG (0.78), SS (0.58) and NoT (0.55) were the most significant contributing traits; similarly, PH (0.10), DoM (0.71) and Pl (0.49) were the important parameters of the second PC.

Phenotypic correlation
Pearson's correlation coefficient was estimated among 10 morphological traits (Table 4)

Discussion
For breeders to begin to produce new varieties in a breeding program, the genetic diversity of plant populations is important. The importance of genetic variation ultimately determines the efficacy of the selection process. In order to obtain new superior varieties, selection is the basis of all breeding activities. Among the 36 rice lines in this study, all the traits demonstrated highly significant (p < 0.05) variation except for GW and Gl. Many researchers have reported the phenotypic variation for different agromorphological characters in rice populations [Hasan et al. 2015b, 22-24]. There was significant difference among 30 homozygous blast resistant lines of rice (MR263 × PS2) with ten morphological traits [25]. A study carried by [24] demonstrated a highly significant divergence among rice lines using 5 quantitative characters, 190 F 3 generation plants resulting from single crosses between local varieties (Bugis and Sriwijaya). In addition, significant variation was recorded among F 5 lines (Tulaipanji × IR64), F 3 lines between two crossess (Tulaipanji × IR64 × PB1460) and (Badshabhog × Swarna sub1) with 11 quantitative traits [22]. High variability of breeding materials contributes to the subsequent generation with better chance of producing superior recombinants and ability to gain the desired character [26,27]. The present study proposed that all the traits recorded higher PCV compared to GCV. This was in line with the studies by Hashemi et al. [28] and oladosu et al. [24] for all the traits observed, who reported that a high influence of the environment on the expression of characters, or higher sensitivity of phenotypes to environment and greater role of phenotype control governing the character. It suggests that the characters were much affected by environment, and selection on the basis of phenotype independent of genotype could be effective for improvement of such traits. Assessment of heritability and genetic advancement is important for selection based on phenotypic expression, according to Johnson et al. [21]. Thus, high heritability with high genetic advance will make the selection based on morphological characteristics effective [32]. In the present study, high heritability with high genetic advance was recorded for the following five traits: namely plant height, number of tillers, filled grain, 1000-grain weight and seed setting rate. It indicates the presence of additive genetic control, which is also in agreement with Mazid et al. [33] and their study in bacterial blight resistant rice genotypes. The findings were also in agreement with the studies reported by Pandey and Anurag [34], Akhtar et al. [35], and Sohrabi et al. [36] for number of filled grain panicle and yield. Thus, direct selection for these characters may be effective in segregating population for improvement of grain yield. The UPGMA dendogram widely grouped the BC 2 F 3 rice lines into four major clusters at a dissimilarity coefficient of 12.33, which indicated a high level of morphological variation in the rice lines. The findings in this study showed an improved approach to group the quantitative traits. Principle component analysis confirmed the existence of broad morphological difference among the genotypes, which stipulated that the total diversity shown can be explained by a few eigenvectors. Based on the morphological traits, 71.13% of the total variation was described by the first three principal components, with PC1 elucidating 39.18% of the variation, PC2 19.98% and PC3 11.97%. An almost parallel finding was also recorded by laslita-Zapico et al. [32] and Tuhina et al. [33], who reported about 69.3% and 72% of total variation, respectively, in all studied traits and recorded high correlation among the them.
Knowledge on the correlation among different traits is crucial to construct an implicit breeding strategy for any crop. Yield characters are quantitatively inherited and influenced by genetic effects and the interaction between genotype and environment. Therefore, direct selection to develop yield could be intricate and laborious due to some selected characters being expressed late in plant development. Thus, indirect selection is much easier and preferable. Accordingly, it is proper to determine and practice the selection of highly correlated characters [34,35]. In the present study, yield had significant positive correlation with number of tillers, 1000-grain weight and seed setting rate. Mazid et al. [29] reported a positive correlation between the 1000-grain weight and grain yield. Furthermore, a significant positive correlation was noted between yield and number of tillers, which was in agreement with the study of Zhou et al. [11]. Principally, the significant phenotypic correlation between traits was possibly due to the pleiotropic effect rather than linkage between the genes influenced by different traits [36]. Negative correlation was also observed between yield and plant height, between plant height and 1000-grain weight, between plant height and filled grain and between plant height and days of maturity, which was also reported by Kumar and Singh [37].

Conclusions
The present study indicated the presence of a considerable level of diversity among 36 BC 2 F 3 rice lines. The highest heritability and genetic advance were found for the plant height, number of tillers, filled grain, 1000-grain weight and seed setting rate, are highlighted as crucial for trait selection for hybridization programs. Furthermore, based on the cluster analyses, the selected blast resistant lines from 5, 6, 7, 20, 21, 23, 24, which produce high yield and have ideal morphological traits, could be used as parents for future breeding programs.