Introduction
Rice (Oryza sativa L.) is the world's most important crop and a primary food source for half of the world's population. It accounts for 35–60% of the caloric intake of three billion Asians (Guyer et al., Reference Guyer, Tuttle, Rouse, Volrath, Johnson, Potter, Gorlach, Goff, Crossland and Ward1998) and 40% of the protein consumed on an average in Asia. Among the two cultivated species, O. sativa, native to Asia, was extensively used in crossing programmes after the identification of the dwarfing gene Dee-geo-woo-gen, which led to the narrowing of the genetic base followed by pre-disposition to biotic and abiotic stresses. Oryza glaberrima, endemic to Africa, is reported to be tolerant to a number of biotic and abiotic stresses and has useful traits that include early maturity, better weed competitiveness, ceremonial and cultural importance (Maji and Shaibu, Reference Maji and Shaibu2010). In India, O. glaberrima is used in limited extent for the genetic improvement of indica rice (Sarla and Swamy, Reference Sarla and Swamy2005). The breeding of high yielding cultivars with wide adaptability is the ultimate aim of plant breeders, and the knowledge of genetic variability for the trait under improvement is of great significance for the success of any plant breeding programme. Heritability and genetic advance are important selection parameters as heritability estimates along with genetic advance are more helpful in predicting the gain under selection. Since high heritability does not always indicate high genetic gain, heritability along with genetic advance should be used in predicting the ultimate effect for selecting superior varieties (Ali et al., Reference Ali, Khan and Asad2002).
Correlation and path analysis determine the association between yield and its components and also bring out the relative importance of their direct and indirect effects, thus proving an understanding of their association with grain yield. Essentially, this kind of analysis could benefit the breeder to choose his selection strategies to improve grain yield. Genetic diversity is also an important tool for a crop improvement programme, as it helps in the development of superior recombinants (Manonmani and Khan, Reference Manonmani and Khan2003). Mahalanobis's D 2 statistics has been used in several crops for identifying diverse parents for hybridization programmes. It is a powerful tool used to quantify the genetic divergence between the accessions and to relate clustering pattern with the geographical origin. Principal component score strategy has been employed for the identification set of accessions which captures maximum genetic diversity of the whole collection (Noirot et al., Reference Noirot, Hamon and Anthony1996; Gireesh et al., Reference Gireesh, Husain, Shivakumar and Satpute2017) and has been successfully used in the germplasm evaluation of crops for understanding the relationship and correlation among the variables studied (Zafar et al., Reference Zafar, Arshad, Ashraf, Mahmood and Abdul2008).
The great diversity of the Oryza genus could serve as a pool of genes for the improvement of cultivated varieties, and in this regard, the African rice would be a good tool for varietal improvement of O. sativa. Thus, the present investigation is carried out with the objective of studying the genetic variability, diversity and trait relationship associations in African rice (O. glaberrima) for yield improvement. The present study has also attempted to identify the set of O. glaberrima accessions which capture maximum genetic diversity using PCS strategy.
Material and methods
Thirty-one O. glaberrima lines received from IRRI (International Rice Research Institute) were sown in dry bed during Kharif 2016 and Kharif 2017 at the ICAR-Indian Institute of Rice Research, Hyderabad. Twenty-seven days old seedlings of each accession were transplanted by adopting a spacing of 20 cm between rows and 15 cm between plants in a Randomized Block Design with two replications. All necessary precautions were adopted to maintain a uniform plant population per replication.
Mean data were collected for seven quantitative characters at the appropriate growth stage of the plant. The characters that were evaluated included days to 50% flowering, plant height (cm), panicle length (cm), number of productive tillers, number of spikelets per panicle, 1000 seed weight (g) and grain yield per plant (g). The mean data after computing for each character was subjected to standard methods of analysis of variance following Panse and Sukhatme (Reference Panse and Sukhatme1985). Correlation coefficients were calculated using the formulae suggested by Falconer (Reference Falconer1964), path analysis by Dewey and Lu (Reference Dewey and Lu1959) and regression analysis through excel. Principal component analysis (PCA) and principal component score were derived using XLSTAT software to reveal the best relationships among traits, while D 2 analysis was performed using INDOSTAT software.
Results
Genetic variability of morphological traits
In the present study, analysis of variance revealed significant differences among the accessions under study for all the seven traits studied (online Supplementary Table S2) for 2 years of evaluation. Days to 50% flowering recorded a general mean value of 117.89 d ranging from 89.8 d (EC 861816) to 138 d (EC 861791) combined over both the seasons (online Supplementary Table S3). Out of all the accessions studied, one accession (EC 861816) was early (<90 d), 17 accessions were medium (101–120 d), 10 accessions were late (111–130 d), while three accessions (EC 861786, 861790 and 861791) were very late (>131 d) flowering accessions. The plant height varied from 91.5 (EC 86186) to 120.7 cm (EC 861785). The mean value for productive tillers was 10.03 with a range from 6.35 (EC 861814) to 12.84 (EC 861794). With respect to the panicle length, longest panicle was of the accession EC861809 (28 cm) while the shortest one was of EC861804 (21 cm). A maximum of 114.9 and a minimum of 67.2 spikelets per panicle were observed in accessions EC861787 and EC861804, respectively. The 1000 seed weight varied from 17.1 g (EC 861805) to 23.4 g (EC 861804) and designated as low (15–20 g) in 16 and medium (21–25 g) in the remaining 15 accessions. Grain yield per plant ranged from 6.70 g (EC 861816) to 14.25 g (EC 861792) with a general mean of 10.18 g.
The genetic variability depicted in the form of box plots (Fig. 1) showed the frequency distribution for seven quantitative traits among 31 accessions of O. glaberrima. The traits, namely, plant height, productive tillers, panicle length and 1000 seed weight, exhibited greater genetic variability in Kharif 2017 compared to Kharif 2016. Though days to 50% flowering exhibited more variability in Kharif 2016, it had an outliner at the bottom with early flowering in Kharif 2017. The variability in terms of spikelets per panicle with outliners was comparable in both the seasons. The grain yield per plant exhibited greater variability in Kharif 2016 than Kharif 2017. With respect to normal probability plot, for Kharif 2016, Fig. 2 depicted normal distribution for plant height, panicle length, spikelets per panicle, 1000 seed weight and grain yield, while the remaining traits, viz., days to 50% flowering and productive tillers had few accessions slightly deviating from the normal distribution indicating variability in the them. For Kharif 2017, normal probability plot depicted normal distribution for all the traits except for plant height, days to 50% flowering and productive tillers.
Box-plots showing the variation of the data from the seven quantitative variables of two seasons evaluated in 31 accessions of O. glaberrima. The upper, median and lower quartiles represent the 75th, 50th and 25th percentiles of the accessions respectively. The vertical lines represent the variation in the population. Dots represent the outliners.
Normal probability plots showing the distribution of O. glaberrima accessions for yield and its attributing traits in (a) Kharif 2016 and (b) Kharif 2017.
Correlation and regression analysis
Correlation analysis among yield and its attributing traits (Table 1) depicted in the form of correlogram (online Supplementary Fig. S1) revealed that grain yield per plant had significant positive correlation with days to 50% flowering (0.479), number of productive tillers (0.470) and spikelets per panicle (0.497), while a negative association was found with 1000 seed weight (−0.294). The trait, 1000 seed weight was found to have a significant negative association with days to 50% flowering and a negative relation with the remaining traits studied. Regression analysis of grain yield with days to 50% flowering, plant height, productive tillers, panicle length, spikelets per panicle and 1000 seed weight provided a regression equation of
$$\eqalign{{\bi Y} = &-{\bf 4}.{\bf 35} + {\bf 0}.{\bf 02}\,({\bf DFF}) + {\bf 0}.{\bf 0033}\,({\bf PH}) + {\bf 0}.{\bf 32}({\bf PT}) \cr &+ {\bf 0}.{\bf 035}({\bf PL}) + {\bf 0}.{\bf 057}\left( {\displaystyle{{\bi S} \over {\bi P}}} \right)-{\bf 0}.{\bf 07}\,({\bf TSW})} $$with R 2 = 0.4702. Regression analysis revealed that yield had a positive relation with days to 50% flowering (0.02), plant height (0.0033), productive tillers (0.035), panicle length (0.035) and spikelets per panicle (0.057), while a negative association with 1000 seed weight (0.07). Coefficient of determination gives information on how much variation of dependent variable was due to the independent variable, which gave a value of R 2 = 0.4702, indicating 47.02% variation of the yield parameter was due to the independent traits.
Correlation coefficient analysis of yield and yield attributing traits in O.glaberrima accessions
*-significance at 5%; **-significance at 1%.
Path coefficient analysis
Path coefficient analysis revealed the trait, number of productive tillers exerting the highest direct positive effect (0.5793) on grain yield per plant followed by spikelets per panicle (0.4195) (online Supplementary Table S4, Fig. 3). The direct positive effects of the remaining characters were low to be considered of any consequence. On the other hand, negative direct effect on grain yield was recorded by days to 50% flowering (−0.0046) and panicle length (−0.0862). Although days to 50% flowering had a negative direct effect on grain yield, its indirect effect through panicle length (0.0002) was positive. The high positive indirect effects on grain yield per plant were of spikelets per panicle via plant height (0.2291) and spikelets per panicle via days to 50% flowering (0.2161).
Path diagram for Grain yield per plant in O. glaberrima accessions.
Principal component analysis
PCA revealed two most informative principal components with Eigen values of 2.643 and 1.189, respectively, which together accounted 54.752% of the total variance for all the characters of both the seasons (Table 2). According to principal component 1, characters such as yield (0.809), days to 50% flowering (0.764) and number of spikelets per panicle (0.669) had relatively higher contributions (37.763%) to the total morphological variability, while the second principal component accounted for 16.988% of the total variation with panicle length (0.541) giving the highest contribution. In Scree plot (online Supplementary Fig. S2), principal component 1 showed 37.76% variability with Eigen value 2.643 which then declined gradually. Accessions were selected based on their principal component scores (online Supplementary Table S5) >1.0. In PC 1, the principal component scores having positive scores ranged from 1.281 (EC861784) to 3.179 (EC 861792), while in PC 2, the positive value of the components ranged from 1.081 (EC 861804) to 2.812 (EC861820) and a cumulative of 56.007% was accounted from the accessions EC 861813, EC861816, EC861792 and EC861804 contributing towards the total diversity.
Eigen values, contribution of variability and factor loading for the principal component axis in O. glaberrima accessions
Cluster analysis
In the present study, heatmap (online Supplementary Fig. S4) revealed spikelets per panicle exhibiting greater variation among the accessions followed by panicle length and days to 50% flowering with blue and yellow colours corresponding to low and high diversity for expressed traits, respectively, while green representing median levels of expression. The 31 accessions were grouped into six clusters based on D 2 values using the Tocher method. The distribution of accessions into various clusters is displayed in Table 3, online Supplementary Fig S3. Out of the six clusters, cluster I was the largest comprising of 25 accessions followed by cluster II with two accessions, i.e. EC 861786, 861791. Clusters III, IV, V and VI had only one accession each, i.e. EC 861804 in cluster III, EC 861805 in cluster IV, EC 861809 in cluster V and EC 861816 in cluster VI. The average intra and inter cluster D 2 values can be computed from the cluster diagram where the statistical distances among the 31 accessions were exhibited (online Supplementary Table S6). Intra cluster D 2 values ranged from zero to 1.93 with maximum distance in cluster I (5.05), followed by cluster III (1.93). From the inter cluster D 2 values of the six clusters, highest divergence was noticed between cluster II and VI (48.15) followed by cluster II and V (34.04) while the lowest was noticed between cluster I and III (9.88). The cluster means for each of 10 characters (online Supplementary Table S7) indicated that the cluster mean for days to 50% flowering was highest in cluster II (137.63) and the lowest in cluster VI (89.75). Plant height was highest in cluster II (111.08 cm) and lowest in cluster VI (91.50 cm). Cluster V recorded the highest number of productive tillers per plant (11.15) and the lowest was in cluster VI (7.00). Cluster V was characterized by longest panicle length (28.00 cm) while the shortest was recorded in cluster III (20.95 cm). The number of spikelets per panicle was highest in cluster I (92.48) and the lowest number was noticed in cluster II (86.30). Highest 1000 grain weight was recorded in cluster III (23.50 g) while the lowest in cluster IV (17.00 g). Highest grain yield per plant was associated with cluster IV (12.10 g), while in cluster III it was the lowest (6.50 g). It is observed that cluster V had many of the desirable means for several characters and with respect to relative contribution of different characters to the genetic diversity (online Supplementary Table S8), the contribution of days to 50% flowering was highest (43.01%) followed by yield per plant (12.90%), number of productive tillers (10.75%), panicle length (9.89%), plant height (8.60%), 1000 seed weight (8.17%) and spikelets per panicle (6.66%). The characters days to 50% flowering, yield per plant, productive tillers along with panicle length contributed 76.55% towards total divergence.
Grouping of 31 accessions of O. glaberrima into different clusters based on yield and yield contributing traits by D 2 analysis
Discussion
Rice (O. sativa L.) is the major food crop of India and being the primary centre of origin, it possesses huge diversity of both wild and cultivated species. It provides food to over 75% of the Asian population and more than three billion of the world population. Genetic variability is a measure of the tendency of individual plants in a population to vary from one another due to the differences in the genetic material. Presence of genetic variability is essential for the improvement of crops species as it provides an opportunity for the breeders to develop new varieties and hybrids. In the present study, the genetic variability was more in Kharif 2016 for days to 50% flowering and grain yield per plant, while all the remaining traits had higher variability in Kharif 2017 determining their usefulness for planning and execution of a successful breeding programme. Three accessions, viz., EC861794, EC861813 and EC861809, were identified as outliners in box plots for grain yield, number of productive tillers, spikelets per panicle. Hence, their utilization in combination breeding may help in generating high yielding varieties by pyramiding all the favourable genes.
Correlation studies are useful in understanding the association between grain yield and different traits (Dixet and Dubey, Reference Dixet and Dubey1984), enabling plant breeders to select accessions possessing desirable traits that are related to grain yield. Since grain yield per plant was positively correlated with days to 50% flowering, plant height, number of productive tillers, panicle length and spikelets per panicle, selection for these traits could be considered as the criteria for higher grain yield, as they were mutually and directly associated. The 1000 seed weight was found to be negatively correlated with yield per plant which would hinder the expression of the plant yield. Thus, this trait may not be rewarding if selected for enhancing the grain yield as reported by Singh et al. (Reference Singh, Babu, Kumar and Mehandi2014) while working with exotic germplasm of upland rice (O. sativa L.). On the other hand, regression analysis revealed that yield had a positive relation with all traits except for 1000 seed weight. It can be concluded that correlation analysis is supported by regression statistics with days to 50% flowering, plant height, number of productive tillers, panicle length and spikelets per panicle being the important traits, which can be considered for breeding programmes to get enhanced yield in O. glaberrima accessions.
As there are many factors influencing yield, selection based on simple correlation without considering the interaction between different component characters can be deceptive. Therefore, genotypic correlation was partitioned into direct and indirect effects through path coefficient analysis for better interpretation of cause and effect relationship among the traits (Wright, Reference Wright1921). The traits, number of productive tillers and spikelets per panicle exerted the highest direct positive effect on grain yield per plant indicating that selection for these characters is likely to bring about an overall improvement in grain yield directly as suggested by Madakemohekar et al. (Reference Madakemohekar, Mishra, Chavan and Bornare2015) in rice (O. sativa L.), while days to 50% flowering and panicle length had shown a negative direct effect on the grain yield suggesting that selection for these traits would hinder the increase in grain yield. The indirect effects of days to 50% flowering via spikelets per panicle, productive tillers and plant height via spikelets per panicle on grain yield per plant were positive and high, pointing the indirect selection for these traits would be beneficial in increasing the grain yield per plant in rice breeding programmes.
PCA analyses the relationship among accessions for characterization and assessing the germplasm diversity. PCA explained the genetic diversity of the O. glaberrima accessions with two principal components having Eigen values of 2.643 and 1.189 accounting for 54.752% of the total variance. Similar results were reported by Worede et al. (Reference Worede, Sreewongchai, Phumichai and Sripichitt2014) for 61.2% of the total variability using the first and second PCs for 24 rice genotypes. Selection of traits via yield per plant, days to 50% flowering, spikelets per panicle and panicle length lying in these two principal components would be beneficial in contributing to the total morphological diversity. With respect to the accessions, EC 861813, EC861816, EC861792 and EC861804 having high principal component scores contributed 56.007% towards the total diversity. Hence, utilizing these accessions as parents would be rewarding in rice breeding programmes.
Heat maps show the relatedness of accessions and traits under study based on their Euclidean distances, which depicted most of the variation incurring through panicle length and days to 50% flowering. D 2 analysis grouped the accessions into six clusters with cluster I comprising of 25 accessions indicating close relatedness of the accessions. Clusters III, IV, V and VI had only one accession indicating a high degree of heterogeneity which may be directly utilized as parents in future hybridization programmes to combine desirable characters and hybridization between accessions belonging to different clusters, viz., cluster I and V, cluster II and V would certainly be rewarding. Similar results of grouping the accessions into six clusters was reported by Chouhan et al. (Reference Chouhan, Singh, Singh, Singh, Singh and Singh2015) while studying 35 wild rice accessions using D 2 analysis for 14 traits. The characters days to 50% flowering followed by yield per plant contributed highly to the genetic diversity which should be given importance during hybridization and selection in segregating populations by choosing plants with early duration and more grain yield per plant. The groupings of D 2 hierarchical cluster analysis exhibited a similar contribution of characters to the genetic diversity to that produced using PCA analysis.
Conclusion
The present study revealed a high level of variability for days to 50% flowering and grain yield per plant in Kharif 2016, than plant height, productive tillers, panicle length and 1000 seed weight in Kharif 2017. There were direct positive associations between yield per plant and days to 50% flowering, plant height, panicle length, number of productive tillers, spikelets per panicle, selection for which would be effective to enhance the yield potential. The trait, number of productive tillers exerted a highest direct positive effect on grain yield per plant, while a negative direct effect on grain yield was recorded by days to 50% flowering and panicle length. PCA identified yield per plant, days to 50% flowering, spikelets per panicle and panicle length in different principal components similar to D 2 statistics playing a prominent role in classifying the variation existing in the germplasm accessions. Thus, the results of the present study revealed high level of genetic variation existing in the population along with the traits contributing to this diversity, which can find immense applications in rice improvement programme. Findings from the present investigation will help in utilizing the African germplasm for genetic improvement of indica rice varieties.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1479262119000327
Acknowledgements
The authors greatly acknowledge the support provided by the ICAR-Indian Institute of Rice Research, Hyderabad and MFPI-Quality control Lab, PJTSAU, Hyderabad for providing the resources for conducting the experiment.
