Genetic variability studies for tuber yield and yield attributes in Ethiopian released potato (Solanum tuberosum L.) varieties

Information on the extent of genetic variability and association among quantitative traits are vital for any crop improvement program and the development of suitable selection strategies. Limited research has been carried out thus far on potato genetic variability and trait association. This study on genetic variability and association among quantitative traits was conducted to assess the extent of genetic variability among yield and agronomic traits to identify superior varieties for the breeding program. To this effect, 20 improved varieties and a local cultivar were planted at two locations in central Ethiopia during the main cropping season of 2017/18 in a randomized complete block design using three replications. Analysis of variance of tuber yield and yield traits at each location and over locations, revealed the existence of highly significant (P < 0.01) differences among varieties in all agronomic and yield traits. Phenotypic coefficient of variation values ranged from 0.75% (specific gravity) to 32.22% (total starch yield) while the genotypic coefficient of variation values ranged between 0.70% (specific gravity) to 30.22% (total starch yield). Maximum difference between phenotypic and genotypic coefficient of variation values were noted for stem number, average tuber number, average tuber weight, number of leaves per plant and tuber yield. Hence, these traits are substantially influenced by the physiological status of the seed tuber at planting and by the environment, post emergence. Range of variability for most of the traits was high, indicating ample scope for selection and improvement in these traits. The estimated values for broad sense heritability and genetic advance, as percent of mean, ranged from 33.52% to 98.66% and 1.35% to 58.26%, respectively. All the traits had high heritability values, except average tuber number per hill, days to physiological maturity, average tuber weight and number of leaves per plant with moderate heritability values.


INTRODUCTION
programs (Johnson, Robinson & Comstock, 1955). Estimates of heritability based on growing potato genotypes at multiple locations for several years will support potato breeders to decide the breeding strategy that should be followed. Mishra et al. (2006) noticed that high heritability associated with high genetic advance would be used as a clue in most selection programs. Performing principal component analysis on trait of the varieties or populations displaying variability helps to reduce any redundancy in the component variables (Placide et al., 2015).
The national potato research program in Ethiopia needs to improve the crop's yield potential, important agronomic traits as well disease resistance, and quality traits that are key to meeting consumer preferences. Moreover, knowledge on the degree of genetic variability present among genotypes and the association of quantitative characters with yield is vital for any crop improvement program and also to develop suitable selection strategies (Ene et al., 2016). On the other hand, the estimation of genetic variances helps plant breeders to choose the most efficient breeding design for improving crops with the existing resources (Oloyede-Kamiyo, Ajala & Akoroda, 2014). Such information is scanty owing to the limited work done by the Ethiopian potato breeding program within the existing genetic pool in the country. Therefore, it is necessary to study genetic variability in the yield potential, disease tolerance, and other important traits. The present study was carried to investigate and estimate the nature and extent of variability in yield and agronomic traits among 20 released varieties and one local cultivar and thereby identify superior varieties for a breeding program.

Experimental materials and sites
The present study was carried out at two locations in the central highlands of Ethiopia, Holetta Research Centre and Adaberga sub-station research fields for a single cropping season (Table 1). The experiment consisted of 20 released varieties and one local potato cultivar (Table S1). The experiment was laid out in a randomized complete block design replicated three times. Analysis of variance for each location and across locations was done using SAS software version 9.3 (SAS Institute, Cary, NC, USA) as per the procedure indicated for the design using a general linear model (GLM) (Gomez & Gomez, 1984).

Data collection and measurement for the traits
Data collection and measurement of yield and agronomic traits (days to 50% flowering, days to physiological maturity, number of leaves per plant, plant height, stem number per hill, average tuber number per hill, average tuber weight, total tuber yield, marketable tuber yield, specific gravity, dry matter content, starch content percentage and total starch yield) were under taken based on the procedures indicated by Lemma, Wassu & Tesfaye (2020).

Statistical methods
For each quantitative attribute, phenotypic and genotypic variability were estimated using variances and coefficient of variations using the procedure suggested by Burton & de Vane (1953).

Genotypic variance ðs
where σ 2 g = genotypic variance M g = mean square of genotype M e = mean square of error; r = number of replications; Phenotypic Variance (σ 2 p ) = σ 2 g + σ 2 e where, σ 2 g = Genotypic variance; σ 2 e = Environmental variance; σ 2 p = phenotypic variance where PCV = Phenotypic coefficient of variation GCV = Genotypic coefficient of variation x = population mean of the character being evaluated. PCV and GCV values were classified as low, moderate, and high according to Sivasubramanian & Menon (1973) as 0 up to 10% = Low, 10 up to 20 = Moderate and above 20 = High.

Heritability and genetic advance
Broad sense heritability values were estimated for each location using the formula adopted by Falconer & Mackay (1996) as follows where H 2 is broad-sense heritability, σ 2 p-phenotypic variance σ 2 g-genotypic variance The heritability percentage was categorized as low, moderate, and high, as suggested by Robinson, Comstock & Harvey (1949):-0 to 30% ¼ Low; 30% to 60% ¼ Moderate; >60% ¼ High

Expected genetic advance under selection
Genetic advance in absolute units (GA) and percent of the mean (GAM), assuming selection of superior 5% of the genotypes, were estimated for each location based on the methods illustrated by Johnson, Robinson & Comstock (1955): where, GA = Genetic advance SDp = Phenotypic standard deviation on mean basis; H 2 = Heritability in the broad sense. K = the standardized selection differential at 5% selection intensity (K = 2.063).

Genetic advance as percent of mean
Genetic advance as percent of mean was estimated as follows: where GAM = Genetic advance as percent of mean The GAM values were grouped in to low, moderate, and high categories, as suggested by Johnson, Robinson & Comstock (1955):

Estimates of variability components over locations
The computation of genotypic and phenotypic variance across locations considered the expected mean squares from combined analysis of variance and the σ 2 P, σ 2 G and H 2 = phenotypic variance, genotypic variance and heritability in broad sense were determined as indicated in Table 2 as proposed by Jalal & Ahmad (2012). Principal component analysis (PCA) was conducted to reduce the observed agronomic and yield traits as suggested by Amy & Pritts (1991). The study considered the main component of maximum eigenvalue of up to 1.0 to estimate the variability in potato varieties studied.

RESULTS
Results of the variance analysis indicated adequate variability among varieties over all the traits considered (Table 3). The analyzed varieties were widely variable over the analyzed characteristics. Tuber yield varied significantly among improved varieties and local cultivars (13.8 to 32.8 t ha −1 ). The maximum mean values of tuber yield, specific gravity, dry matter content, starch content and total starch yield were recorded from the variety Belete while the lowest mean value of all these traits was recorded from the variety Menagesha (Table 4).
The results of estimates of variability components (phenotypic and genotypic coefficient of variations, heritability, and genetic advance) for each location are presented in Table 5. The estimates of the variability of components across sites are shown in Table 6.

Phenotypic and genotypic variation
At Holetta, the estimated phenotypic (PCV) and genotypic (GCV) coefficient of variations of specific gravity and total starch yield of the different potato genotypes studied ranged between (0.71% to 35.23% and 0.62% to 34.51%), respectively. At Adaberga the PCV and GCV values of specific gravity and total starch yield ranged between (0.91% to 32.75% and 0.83% to 31.97%), respectively ( Table 5). The over locations PCV values range from 0.75% (specific gravity) to 32.22% (total starch yield) while the GCV values ranged Table 2 Estimation of variance components and broad sense heritability on combined analysis of variance over locations basis.

Genotypic parameter Symbol Determination method
Error variance Broad sense heritability H 2 σ 2 G/σ 2 P Note: σ 2 E, error variance; σ 2 L, location variance; σ 2 G, genotypic variance; σ 2 GL, genotype × location interaction variance; σ 2 P, phenotypic variance; H 2 , heritability in broad sense; MSE, mean square error; MSL, mean square location; MSG, mean square genotype; MSGL, mean square genotype × location interaction; and rl, replication by location. Note: * , ** , significant at P = 0.05 and P < 0.01, respectively. Rep = replication, CV (%) = coefficient of variation in percent, numbers in the parenthesis are degrees of freedom. from 0.70% (specific gravity) to 30.22% (total starch yield) ( Table 6). The results of the phenotypic variance were in general higher than the genotypic variance for all characters studied. The difference between PCV and GCV was the most pronounced for the traits stem number, average tuber number, average tuber weight, number of leaves per plant, tuber yield. Therefore, we speculated that these traits are substantially influenced by the growing environments.

Estimates of heritability and genetic advance
The estimated values of H 2 and GAM ranged from 33.52% (average tuber number per hill) to 98.66% (days to 50% flowering) and 1.35% (specific gravity) to 58.26% (total starch yield), respectively (Table 6). All the traits had high heritability values but average tuber number per hill, days to physiological maturity, average tuber weight, and the number of leaves per plant. Highest GAM was recorded for total starch yield, starch content percent, total and marketable tuber yields, plant height, and the number of stems Note: DF, days to 50% flowering; DM, days to physiological maturity; NLP, number of leaves per plant, PH, plant height (cm); SN, stem number per hill; ATN, average tuber number per hill; ATW, average tuber weight(g/tuber); TTY, total tuber yield (t ha −1 ); MTY, marketable tuber yield (t ha −1 ); SG, specific gravity (gcm −3 ); DMC, dry matter content (%); SC, starch content percent (g 100 g −1 ); TSY, total starch yield (t ha −1 ).
recorded for days to physiological maturity, the number of leaves per plant, average tuber number per hill, and average tuber weight per plant, suggesting non-additive gene action in their control: hence complex breeding methods might be recommended improving potato yield through these traits.

Principal component analysis
The principal component (PC) analysis was used to combine the observed traits into four main components that have eigenvalues above 1.0 and was able to explain the tested variability at 87.53% (Table 7). PC1 with the eigenvalue 5.7 explained 47.53% of the total variance, PC2 with the eigenvalue 2.1 contributed to 17.73% of the total variance, PC3 with the eigenvalue of 1.6 contributed to 13.57% of the total variance and PC4 with the eigenvalue 1.0 contributed to 8.70% among the 20 potato varieties and one local cultivar. The first principal component is strongly positively correlated with plant height, total tuber yield and total starch yield. The cumulative variance of 87.53% by the first four axes indicated that total tuber yield, total starch yield ton per hectare, and plant height contributed to the larger share of the observed variations among potato varieties and could effectively be used for selection. Other phenotypes, such as the number of "days to 50% flowering", "average tuber number per hill", and "marketable tuber yield ton per hectare" were also important, as they negatively affect the yield (Table 7).

DISCUSSION
The observed genetic variability among the evaluated varieties in the present study indicated the presence of breeding lines that can be used as parental materials for the potato improvement program (Table 4). Four varieties (Gudene, Gabissa, Gorebella, and

Phenotypic and genotypic variation
The range of PCV and GCV variability for most of the characters were high, indicating the presence of sufficient scope for selection and improvement in these characters, including tuber yield t ha −1 . Similarly, maximum PCV and GCV values were reported by Abraham (2013) and Fekadu, Petros & Zelleke (2013) and Singh (2008) for tuber yield and other yield-related traits. These results agreed with Tripura et al. (2016), who reported high PCV and GCV values for tuber yield and additional yield-related traits as average tuber weight, tuber number, tuber breadth, and tuber yield in potato. Low PCV and GCV values were recorded for some tuber quality traits viz., specific gravity and dry matter content which is due to the fact that most tuber quality traits like tuber specific gravity and dry matter content are less influenced by environments . Likewise, Chepkoech et al. (2018) reported moderate PCV values and low GCV values in plant height, stem number, tuber number, and tuber weight traits. Low to moderate PCV and GCV values were also reported for tuber length (mm), single tuber weight (g), plant height (cm), branch number, lateral leaflet, and the number of tubers per plant.

Estimates of heritability and genetic advance
In the present study, heritability estimates were high for most of the studied traits as categorized (Low <30%; Moderate 30-60%; High >60%) by Robinson, Comstock & Harvey (1949) and Johnson, Robinson & Comstock (1955). The presence of high heritability in most traits indicates a lower environmental influence. Similar results were reported by Panja et al. (2016). Johnson, Robinson & Comstock (1955) categorized GAM values as Low (0-10), Moderate (10-20), and High (≥20). In the present study, results of analysis of GAM of all the studied traits had high to moderate values except specific gravity and days to physiological maturity that exhibited low GAM (1.35 and 3.83), respectively. High value of H 2 and GAM indicated the existence of genetic variation in potato varieties which could be an opportunity for further improvement of potato. Similarly, Tripura et al. (2016) reported high heritability values coupled with high genetic advance for an average tuber weight plant −1 , single tuber weight, and tuber breadth. Similar reports for total tuber yield (kg per plot) were reported by Rangare & Rangare (2017) and Panigrahi et al. (2017). Likewise, Ozturk & Yildirim (2014) published moderate to high heritability for plant height, leaf width, leaf length, single tuber weight, plant yield, and starch content in potato. Low H 2 and GA values were reported by Fekadu, Petros & Zelleke (2013) for days to emergence, days to flowering, days to maturity, stem number, tuber yield, number of tubers, harvest index, medium tuber percentage, and biomass yield in potato.

Principal component analysis
The results of principal component analysis (PCA) for the studied traits are presented in Table 7. In the present study, the first three traits (plant height, total tuber yield and total starch yield) explained 47.53% of the variation. The result indicated that the cumulative variance of 87.53% was explained by the first four axes (Table 7). Afuape, Okocha & Nijoku (2011) reported a cumulative variance of 76.00% for the first three axes in evaluating twenty-one sweet potato genotypes. Koussao et al. (2014) identified four principal components, which accounted for 67.22% of the total variation among the accessions. Placide et al. (2015) also used PCA to study the fifty-four sweet potato genotypes and found that the first seven PCs explain the cumulative variance of 77.83%. Rahanjeng & Rahayuningsih (2017) reported 79% of variability using sixty-two sweet potato accessions. In our study the first four principal components explained 87.53% of the variability. This value is within the range mentioned in earlier studies.

CONCLUSIONS
In general, this study found diverse genetic variability estimates among potato varieties studied. The agronomic characters of potato varieties showed high genotypic coefficient of variation and phenotypic coefficient of variation values, high estimation of broad-sense heritability values for most traits some coupled with genetic advance as percent of mean. Hence, this information and the identified varieties and traits could be used for future potato breeding programs in the country.

ADDITIONAL INFORMATION AND DECLARATIONS Funding
The authors received no funding for this work.