Genetic variability of panicle architecture in traditional nger millets from Koraput valley of Eastern Ghats for crop improvement

Panicle phenotyping is most demanding targets in crop breeding programs as panicle is an important plant parts that inuences the grain yield. Diversity of panicle traits were analyzed in 20 traditional nger millet genotypes and three hybrid varieties from Koraput valley of Eastern Ghats. Signicant difference of panicle characteristics like panicle length, panicle number, panicle weight, ag leaf area, panicle angle and grain yield were observed in experimented nger millet genotypes. In regards to principal component analysis, the rst two axis of principal component obtained 52.97% of total variation and reected huge difference between explored genotypes. Highest positive loading was observed for leaf weight followed by panicle number, panicle weight and panicle length and leaf area and are the major determinant for phenotypic variability. All the studied traits showed maximized phenotypic coecient of variation (PCV) over genotypic variation (GCV) and lower differences found among them. The genotypes showed maximum broad-sense of heritability value for grain yield (82.27%) and panicle weight (75.96%) compared to other traits. In addition, genetic advance as mean% (GAM) ranged from 11.01–29.26% and high GAM was recorded for panicle angle, leaf area, panicle weight, panicle number and yield. It revealed that these characters can be used as selection criteria in crop improvement program for improving grain yield. Traditional nger millet genotypes such as Bhadi, Murda, Telgu and Chilli recorded superior panicle traits and Limca and Kalia showed better grain yield at par with the hybrids, which can be utilized in future crop improvement program. The nger millet genotypes are grown in the experiment garden of Regional Center of Swaminathan Research Foundation, Koraput during Kharif season in three replications with randomized block design. Healthy seeds of all the genotypes were selected shown in the nursery bed. Experiment eld was prepared by adding organic fertilizer in the form of farmyard manure. After 25 days of growth 3 seedlings per hill was transplanted in the area of one m − 2 with 20 cm x 15 cm spacing between the hill. The plants are regularly irrigated with running water. The maximum light intensity, air temperature and humidity were recorded daily basis and were 1310 ± 60 µmol m − 2 s − 1 , 33.5 ± 3ºC and 60–70%, respectively. samples laboratory and grain yield each genotype expressed


Introduction
Finger millet (Eleusine coracana (L.) Gaertn) is most valuable cereals generally grown on arid and semi-arid regions of the globe under rainfed agro-climatic conditions (Chandra et al. 2016). India is the major producer of nger millet which is cultivated on 1.27 million hectares land brings about 1.89 million tons of yearly production (Sood et al. 2019). Special aspects of millets like the capacity to adjust under extremely unfavorable climatic condition, minimal agroinput requirement and remarkable nutritional value label these crops for future food and nutrition security (Gupta et al. 2017). The crop farming related to various eco-geographical climatic conditions worldwide determines higher genetic variations and perhaps fundamental basis for the future breeding programs (Ceasar et al. 2018). However, the genomic assets accessible for nger millet are limited as contrasted to vital cereals such as rice, wheat and maize which hamper the improvement of this valuable crop (Ceasar et al. 2018). Improvement of nger millet productivity remains the rst concern research in majority of the breeding programs intending for climate resilient agriculture. Crop yield is the most important multifaceted traits and is the nal outcome of combination of different traits like panicle number, panicle length, grain number and weight of grains etc. (Kumari and Singh 2016;Patil et al. 2018). Panicle characteristics in nger millet are regarded as most intuitive part that controls the quality and yield of nger millet (Wu et al. 2016;Lei et al. 2018). Therefore, panicle architecture is an important component and for the most part makes biggest contribution to grain yield (Zhou et al. 2014;Zhao et al. 2016;Patil et al. 2018). Many genes and QTL responsible for panicle traits and yield associated traits has been reported in nger millet (Saha et al. 2016;Ceasar et al. 2018;Sood et al. 2019). However, in compared to other cereals, relatively less information is reported in nger millets in relation to panicle phenotyping and the relationship between panicle characteristics and grain yield is not yet been explored.
Identifying superior germplasm for increasing productivity of nger millet will assist to acquire food and nutritional security (Ceaser et al. 2018). Earlier studies have been attempted to recognize markers linked to important agronomic and nutritional traits (Mutch et al. 2005;Babu et al. 2014;Kumar et al. 2015;Gimode et al. 2016). However, breeding of nger millet with improved qualities is presently restrained by the absence of e ciently characterized germplasms at the agro-morphological levels (Rajendran et al. 2016;Sood et al. 2019). Previously some attempts has been explored to utilize different morphological and agronomic traits for analyzing genetic diversity of nger millet accessions of Africa and Asia (Dida et al. 2008), global nger millet accessions (Babu et al. 2014;Ramakrishnan et al. 2016;Wakista et al. 2017) and genotypes from India (Babu et al. 2018). The genotypes having superior agronomic traits can be utilized in breeding program. Though India holds good numbers of germplasms of nger millet and the majority of the germplasms are collected from the Andhra Pradesh, Karnataka and Tamil Nadu etc. However, very little information is available on genetic variation of traditional nger millet genotypes of Eastern Ghats, particularly from Koraput.
Koraput valley of Eastern Ghats of India is one of the agro-biodiversity hot spots in India and home to a huge number of traditional nger millets (Pradhan et al. 2019). This region with varied agro-climatic conditions surrounded with thick forest and steep mountainous ranges favorable for millet cultivation by local tribes (Panda et al. 2020b). Nowadays due to modern farming practices, these valuable genetic assets are being continuously eroded. The traditional varieties grown by the farmers are likely to be lost shortly if measures are not taken to protect the valuable genetic resources. Recently, some research highlighted the cultivation practice, nutritional and genetic potentiality of nger millets from Koraput (Pradhan et al. 2019;Panda et al. 2020a, b). However, there is limited phenotypic knowledge in regards to panicle phenotyping and yield related traits. Therefore, the objectives of the current study are to elucidate the genetic variation of panicle traits in underutilized nger millet genotypes from Koraput valley and to elucidate the genetic parameters such as heritability and genetic advance which will be useful for future breeding program.

Materials And Methods
Experimental material and growth condition Twenty different traditional nger millet genotypes and three hybrid varieties (GPU66, ML365 and Bhairabi) popularly cultivated from Koraput valley of Eastern Ghats were selected for the study. The details of nger millet genotypes with their maturity duration and special characteristics are presented in Table 1. The nger millet genotypes are grown in the experiment garden of Regional Center of Swaminathan Research Foundation, Koraput during Kharif season in three replications with randomized block design. Healthy seeds of all the genotypes were selected shown in the nursery bed. Experiment eld was prepared by adding organic fertilizer in the form of farmyard manure. After 25 days of growth 3 seedlings per hill was transplanted in the area of one m − 2 with 20 cm x 15 cm spacing between the hill. The plants are regularly irrigated with running water. The maximum light intensity, air temperature and humidity were recorded daily basis and were 1310 ± 60 µmol m − 2 s − 1 , 33.5 ± 3ºC and 60-70%, respectively.

Phenotypic Characterization
Three different panicles were collected from healthy seedlings during the maturity stage (20th to 30th December 2019). The panicles were cut down through a sharp scissor about 1 cm below from rst leaf during collection. The panicle samples were photographed by using a DSLR camera and brought to the laboratory for measurement of panicle traits.
Panicle numbers of each plant were counting by total branch number emerged per hill. Angle between leaf and panicles of individual sample was determined by using a protractor during collection time. Leaf area was determined by measuring length and width of leaf through a measuring scale and further calculated following Yoshida et al. (1976) formula [Leaf Area = length×width×0.67 cm 2 ]. Scienti c ruler was used to determine panicle length from 1 cm down of ag leaf to the tip. Dry weight of leaf and panicle were obtained by oven dried (72 ºC) of individual sample around 48 h and subsequent weighing using digital electronic balance. After harvesting of total plot, the grain weight was noted and grain yield of each genotype was expressed in t per ha − 1 .

Measurement Of Genetic Parameters
Genetic variability parameters such as variance at genetic level (σ 2 G ) and variance at phenotyping level (σ 2 P ) were calculated following the formula of Steel et al. (1997). Then phenotypic coe cient of variance (PCV), genotypic coe cient of variation (GCV) and environmental coe cient of variation (ECV) were determined according to Burton and Devane (1953). By taking the PCV and GCV value, heritability (h 2 bs ) was calculated following Falconer and Mackay (1996). Further, as per Johnson et al. (1955) genetic advance (GA) and genetic advance as mean% (GAM) were calculated.

Statistical analysis
Different parameters were recorded triplicate and one-way ANOVA was performed by applying CropStat-7.2 software. Multiple correlation and multivariate analysis (Principal component analysis and Cluster analysis) were carried out by PAST-3 software. Based on the panicle phenotyping parameters, dendrogarm exhibiting Bray-Curtis similarity index among studied nger millets were constructed using UPGMA cluster analysis.

Diversity of panicle traits
Morphological diversity of panicles in studied nger millets were depicted in Fig. 1. Variation of different morphological characteristics like panicle number, panicle length, panicle weight, leaf area, leaf weight, panicle weight and grain yield were shown in Table 2. The summary statistics of various panicles attributes in nger millet genotypes were shown in Table 3. The coe cient of variation (CV) of different panicle traits were ranged from 4.501-25.450% among the nger millet genotypes. The maximal CV % was noticed in panicle angle (25.45%) followed by leaf area (23.60%), panicle number (19.81%), panicle weight (17.21%), grain yield (13.59%), panicle length (10.10%) and lowest CV was observed in leaf weight (4.501%). The ANOVA analysis also clearly showed that remarkable variation between genotypes for studied panicle traits ( Table 3). The value of variance was highest in panicle angle (31.2%) followed by leaf area (8.51%), panicle length (9.0%) and lowest variance was observed in leaf weight (0.003%).  (Table 2). Signi cant (P < 0.05) difference of grain yield were recorded among the studied nger millet genotypes. The hybrid variety 'Bhairabi' showed highest yield 2.72 t ha − 1 , whereas 'Limca' genotype showed the lowest yield of 1.45 t ha − 1 with a mean value of 2.00 t ha − 1 across the genotypes ( Table 2).

Genetic Variability Parameters
Genetic variability of studied morphological traits was shown in Table 4. All the studied traits showed higher phenotypic variation (σ 2 P ) than that of genotypic variation (σ 2 G ). In addition, PCV was > GCV and ECV in studied panicle traits and lower differences found among them (Table 4). The trait showed the largest PCV data for panicle angle (25.34%) next to leaf area (23.15%) and lowest PCV data was obtained for panicle length (9.957%). Similarly, GCV value varied from 7.29-18.97% and panicle angle showed highest GCV (%) accounts for 18.97% (Table 4). The broad sense of heritability (h 2 bs ) varied from 50-82.270% and the highest h 2 bs value was observed for grain yield (82.270%) followed by panicle weight (75.968%) among the nger millets. The highest genetic advance (GA) was recorded for panicle angle (6.424%). The GAM value varied from 11.007-29.269% and maximum GAM was recorded for panicle angle (29.269%) followed by leaf area (26.419%), panicle weight (26.342%), panicle number (23.733%) and yield (22.477%) ( Table 4). Table 4 Genetic variability parameters such as range, mean, standard error (SE), genotypic variation (σ 2 G ), phenotypic variation (σ 2 P ), environmental coe cient of variation (ECV), genotypic coe cient of variation (GCV), phenotypic coe cient of variation (PCV), heritability in broad sense (h 2 bs ), genetic advance (GA) and genetic advance as percentage of the mean (GAM) for different traits in nger millet genotypes of Koraput. Variations of panicle traits among nger millet genotypes were evaluated by multivariate analysis including PCA (Table 5; Fig. 2). All the studied parameters were separated into seven principal components (PC), among which, three PC having Eigen value > 1 (Table 4). In PC 1 the highest positive loading was found for leaf weight (0.588) succeeded by leaf area (0.442), panicle weight (0.424) and panicle length (0.390) with a variance of 38.06%, whereas in PC 2 maximal variability was marked in panicle angle (0.726) and leaf area (0.331). In PC 3, highest positive loading was noted for panicle number (0.792) followed by panicle length (0.432) and panicle angle (0.256) with a variance of 14.28% (Table 4). The initial 2 axis of principal components (PC1 and PC2) captures 52.97% of total variation. The scatter plot was drawn between two principal components represented a clear pattern of grouping of nger millet cultivars into four quarters (Fig. 2).  Relationship among different panicle traits in studied nger millet were carried out by multiple correlation analysis (Fig. 3). Based on the results, panicle length was positively associated with leaf area, leaf weight, panicle weight and yield and negatively related with panicle number and panicle angle (Fig. 3). In addition, on the basis of Pearson's correlation coe cient, panicle number was positively correlated with panicle length (r = 0.997), panicle angle (r = 0.725), leaf area (r = 0.298) and panicle weight (r = 0.532) ( Table 6). Grain yield in nger millet was positively related with panicle length (r = 0.933), panicle angle (r = 0.531), leaf area (r = 0.843), leaf weight (r = 0.439) and panicle weight (r = 0.654) ( Table 6). Cluster analysis based on panicle traits in nger millet genotypes The genotypic relationship among 20 indigenous nger millet genotypes and three hybrid cultivars was evaluated through Bray-Curtis similarity index (Fig. 4). Based on studied morphological traits, nger millet genotypes clustered into 2 large clusters and different sub-clusters. The largest cluster consists of 15 nger millet genotypes with more than 88% similarity which consisting of 13 indigenous genotypes (Dasara, Kalua, Ladu, Muskuri, Biri, Bada, Chilli, Kurkuti, Lala, Limca, Murda, Mami and Bhadi) and 2 hybrid variety (GPU 66 and ML 365). The hybrid variety 'GPU 66' showed more than 95% similarity with 'Kalua' and 'Dasara' and present in a small sub group. Similarly, in another major cluster, seven indigenous nger millet genotypes along with hybrid variety 'Bhairabi' showed more than 90% similarity (Fig. 4).

Discussion
Genetic variability study of germplasm provides fundamental information concerning genetic characteristics of the population based upon which breeding techniques are developed (Umar and Kwon-Ndung 2014). This is the rst report on genotypic variability of panicle traits in traditional nger millet genotypes from Koraput valley of Eastern Ghats. The present investigation analyzed panicle phenotyping in 20 traditional nger millet genotypes and compared with 3 hybrid cultivars (GPU66, ML365 and Bhairabi) popularly cultivated in the locality. The analysis of variance clearly indicated that high signi cant difference of panicle traits viz., panicle length, panicle number, panicle weight, panicle angle and grain yield were recorded across the examined nger millet genotypes. Current ndings also exhibited higher CV% of different panicle traits were observed among the studied nger millet genotypes. Thus, it is indicated that there was su cient variation of panicle characteristics in the studied genotypes, which provide ample scope for selecting superior genotypes and could be utilized in future breeding programs. Similarly, physiological variability in different underutilized nger millets from various parts of India also reported earlier by Panda et al. (2020b), Pradhan et al. (2019) and, Kumari and Singh (2015).
The panicle architecture is a key component which enhances grain productivity and basically controlled by panicle number, panicle length, accompanying branching pattern and grains (Huang et al. 2018;Patil et al. 2018). Certain traditional nger millet genotypes like Bhadi, Mudra, Telgu and Chilli exhibited signi cantly higher panicle number and panicle weight in compassion to rest genotypes that determined superiority of panicle traits. Panicle length was highest in 'Chillikangra' and highest panicle angle was found in 'Dasara' and 'Kalua' genotypes in comparison to rest genotypes. Wang et al. (2008) reported the shape, size and angle of panicle is crucial for arrangement of grains in the panicle and also helpful for grain lling in cereal crops. There were signi cant variability of leaf area and leaf weight observed among nger millet genotypes and maximum leaf area and leaf weight was observed in hybrid variety ML 365. It is generally advocated that larger leaf area increases the photosynthesis capacity and improve the dry matter and ultimately helpful for better panicle growth (Yoshinaga et al. 2013). Yield is most signi cant complex traits in nger millet that is affected by numerous genetic and environmental elements (Ulaganathan and Nirmalakumari 2013). In the current experiment signi cant variability of grain yield was recorded among the nger millet genotypes.
Though the hybrid varieties Bhairabi and ML 365 showed highest yield but some of the traditional genotypes such as 'Limca' and 'Kalia' showed better yield at par with the hybrids. Similar results of morphological variations in nger millets are also reported in indifferent germplasm collections (Shet et al. 2009;Ulaganathan and Nirmalakumari 2013;Patil et al. 2018). To know the relationship of grain yield with other panicle traits, data are subjected to multiple correlation analysis. Results indicated that grain yield in studied nger millet genotypes were positively associated with panicle number, panicle length, leaf weight, leaf area, and panicle angle and panicle weight. These ndings are reliable with the previous literature of Reddy et al. (2013); Ulaganathan and Nirmalakumari (2013) and Wolie et al. (2013) that grain weight and harvest index depend on the shape, size and number of panicles. Kumari and Singh (2016) reported that yield of nger millet relies upon different growth and yield characteristics like lled grain number, panicle number and grain weight. In contrast, it was also reported that variation of grain yield was not necessarily associated with leaf area in cereals like nger millet (Panda et al. 2020b) and rice (Subrahmanium 2000; Panda et al. 2020c).
In the present study, panicle angle, leaf area and panicle weight showed the higher PCV and GCV value indicating these traits are more variable in the germplasm. The PCV value was greater than GCV and ECV for panicle traits and also interesting to note that the variation among GCV and PCV were minimum in all traits. This implies about following traits were least affected by environment and additive gene effects indicating genotypes can be improved based on genotypic values (Reddy et al. 2013). Similar results were also published by Wolie et al. (2013) and Kumari and Singh (2015) in different nger millet accessions. High level of heritability and genetic advance will be helpful for selection in crop improvement (Sahu et al. 2017;Panda et al. 2020c). The genotypes under study showed high broadsense of heritability value for grain yield (82.27%) and panicle weight (75.96%) compared to other traits. It indicates that following traits are less impacted by environment and effective for selection on the basis of phenotypic performance. The following results are consistent with previous report of Reddy et al. (2013) and, Kumari and Singh (2015) that higher value of heritability can be used for selection criteria for nger millet improvement program. In the present study GAM value changed from 11.01-29.26% and high GAM was recorded for panicle angle, leaf area, panicle weight, panicle number and yield. It is suggested that these traits can be used as selection criteria in improving grain yield in crop improvement program.
Principal component analysis reveals the pattern of genotypic variation and useful for identifying the major contributing traits for phenotypic variability (Mishra et al. 2019). In the light of the outcomes, rst two principal components capture 52.97% of total variation and highest positive loading was observed for leaf weight followed by panicle number, panicle weight and panicle length and leaf area. It suggested that these parameters are the major determinant for phenotypic diversity in studied genotypes compared to other traits. Similar results of phenotypic diversity of morpho-physiological traits are also proclaimed in traditional rice landraces (Panda et al. 2020c) and in underutilized millets (Panda et al. 2020b) from Koraput. Based on panicle attributes, genotypic relationship among the studied nger millet genotypes were evaluated by cluster analysis (Fig. 4). Bray-Curtis similarity index classi ed the genotypes in two major clusters and different sub-clusters. The largest cluster consists of 15 genotypes with more than 88% similarity which consisting of 13 indigenous genotypes and two hybrid variety. Higher similarity among the genotypes may be because of their similar origin, ecotype and cultivated only in a particular region. Results also recorded that no duplicate genotypes existed and indicated the considerable genetic difference between the genotypes for panicle traits. Three indigenous nger millet genotypes such as Taya, Kalua and Dasara were having maximal similarity index along with hybrid cultivars GPU 66 and Bhairabi. The genetic variability results in indigenous nger millet genotypes may provide baseline information for conservation and could be used for varietal improvement of future breeding programmes.

Conclusion
The present study revealed existence of huge variability of panicle traits in studied nger millet genotypes, which provide ample scope for selecting superior genotypes for future breeding programs. Principal component analysis revealed that panicle number, panicle length, panicle weight and leaf area are the major contributing traits for phenotypic variability between the examined genotypes. Some traditional nger millet genotypes like Bhadi, Murda, Telgu and Chilli revealed superiority of panicle traits and Limca and Kalia showed better grain yield at par with the hybrids, which can be utilized in the future crop improvement program. Higher level of heritability and genetic advance were observed for panicle angle, leaf area, panicle weight, panicle number and yield, which suggested that these characters may be helpful for selection in nger millet improvement programs. Further research on population structure, QTL mapping and allele diversity in these genotypes are required for use in future breeding program.

Declarations
Con ict of interest: The authors declare that they have no con ict of interest