Genetic divergence among eggplant genotypes under high temperatures

The aim of this study was to estimate the genetic divergence among eggplant genotypes for agronomic traits in order to gather information for the selection of genotypes in eggplant breeding programs for tolerance to high temperatures. Ten traits recommended by the International Board for Plant Genetic Resources were analyzed in 24 genotypes, arranged in a randomized complete block design with four replicates. Data were submitted to analysis of variance (P<0.01) and later to the UPGMA and Tocher grouping methods, using the generalized Mahalanobis distance (D2) as dissimilarity measure. Three and six groups of similarity were obtained, respectively, for the multivariate techniques used, UPGMA and Tocher, with concordance in the grouping of 87.50% of the genotypes. The characters fruit length (34.71%), fruit width (35.96%) and fruit length/width ratio (14.08%) were the main contributors to genetic divergence, explaining 90.72% of total genetic dissimilarity. The genotypes presented considerable genetic variability for all agronomic traits analyzed and can be used in eggplant genetic breeding programs for high temperatures.


Research
Hortic. bras., Brasília, v.37, n.3, July -September 2019 I n Brazil, the area cultivated with eggplant (1550 ha/year) is concentrated mainly in the Center-South region (Boiteux et al., 2016). In the Northeast, where temperatures are relatively high, averaging around 28°C and peaking around 40°C (Ramalho, 2013) crop yields have been unpredictable. This is mainly due to flowering coinciding with warmer periods of the year, increasing the occurrence of malformation and/or fruit abortion. In greenhouse crops, where the internal temperatures are higher than the outside, there is a considerable reduction in crop yield in the region (Valadares et al., 2019ab).
The optimal temperature for crop growth and development is in the range of 22 to 30°C (Adamczewska-Sowińska & Krygier, 2013). When the temperature exceeds 32°C, productivity is drastically reduced (Baswana et al., 2006). Adoption of strategies for evaluation and selection of eggplant genotypes and knowledge of the genetic variability involved in traits of agronomic importance are extremely important for the choice of genotypes to compose eggplant breeding programs for high temperature tolerance.
Genetic divergence studies provide these parameters and allow the correct choice of parents which, when crossed, result in high heterotic effect on progenies, maximizing the chances of obtaining superior genotypes in segregating generations (Rotili et al., 2012). These genotypes can be obtained by biometric techniques based on quantification of heterosis or by predictive processes (Nardino et al., 2017).
Among the biometric techniques are diallel analyzes, which generate information about the specific combining ability and heterosis manifested in hybrids and in the prediction of genetic divergence, also keeping in mind VALADARES

ABSTRACT
The aim of this study was to estimate the genetic divergence among eggplant genotypes for agronomic traits in order to gather information for the selection of genotypes in eggplant breeding programs for tolerance to high temperatures. Ten traits recommended by the International Board for Plant Genetic Resources were analyzed in 24 genotypes, arranged in a randomized complete block design with four replicates. Data were submitted to analysis of variance (P<0.01) and later to the UPGMA and Tocher grouping methods, using the generalized Mahalanobis distance (D 2 ) as dissimilarity measure. Three and six groups of similarity were obtained, respectively, for the multivariate techniques used, UPGMA and Tocher, with concordance in the grouping of 87.50% of the genotypes. The characters fruit length (34.71%), fruit width (35.96%) and fruit length/width ratio (14.08%) were the main contributors to genetic divergence, explaining 90.72% of total genetic dissimilarity. The genotypes presented considerable genetic variability for all agronomic traits analyzed and can be used in eggplant genetic breeding programs for high temperatures.
Agglomerative methods (Cruz et al., 2012) seek to genetically discriminate individuals and allow them to be separated into groups by analyzing a set of characters inherent to each individual, grouping them by some classification criteria, so that there is homogeneity within each group and heterogeneity between them. They also basically involve two stages, the first refers to the estimation of a similarity or dissimilarity measure and the second refers to the adoption of a grouping technique.
As dissimilarity measures, we can point out the Euclidean distance, the average Euclidean distance, the average squared Euclidean distance, the weighted distance and the generalized Mahalanobis distance (D 2 ) (Cruz et al., 2012(Cruz et al., , 2014. Genotype grouping can be done by optimization and hierarchical clustering methods. Among the optimization clustering methods are the modified Tocher and Tocher (Vasconcelos et al., 2007;Cruz et al., 2014). Hierarchical clustering methods include the methods of the nearest neighbor, the farthest neighbor, UPGMA (Unweighted Pair-Group Method using Arithmetic Averages), the centroid, the median (or WPGMC), and the Ward's minimum variance (Cruz et al., 2012).
Finally, we can adopt the cophenetic correlation analysis to increase the reliability of the conclusions regarding interpretation based on dendrograms. This establishes a correlation between the similarity or dissimilarity matrix with the generated dendrogram, i.e., compares the actual distances obtained between the accessions with the distances graphically represented (Kopp et al., 2007). The higher the correlation value, the smaller the distortion caused by grouping.
Given the above, the present work aimed to estimate genetic divergence between eggplant genotypes for agronomic traits, aiming to generate information for the choice of genotypes in eggplant breeding programs for high temperature tolerance.

MATERIAL AND METHODS
The experiment was conducted between May and September 2016 at Universidade Federal Rural de Pernambuco (UFRPE), Recife-PE.
Seeds were sown in 128-cell expanded polystyrene trays filled with inert substrate (sifted coconut powder). Trays were kept in greenhouse in the hydroponic system by subirrigation until reaching the point for transplantation, plantlets with three definite leaves. Seedlings were individually transplanted to 5 L pots, containing inert substrate (coconut powder), spaced 1.75 m between rows and 0.60 m between plants.
Plants were cultivated in open hydroponics with substrate, under a 30 m long, 14 m wide, 3 m ceiling height arch, with 50% shading side screens and roof covered with a low-density polyethylene film, 150 micrometers thick.
Mineral nutrition and water requirement of plants were supplied by balanced nutrient solution at each plant development stage. A drip irrigation system was used with 2 L h -1 emitter, automatically controlled by a digital timer, with irrigation amounts and duration adjusted according to environmental conditions of the region and the amount of nutrient solution absorbed by the plants.
Throughout the experiment period, relative air temperature (average, maximum and minimum) and relative air humidity were recorded using a HOBO mini datalogger. The environmental conditions in which the experiment was performed are characterized by high temperatures, since in all phenological phases temperatures exceeded the optimum range of the culture.
Eighteen eggplant accessions from the Embrapa Hortaliças' germplasm bank and six commercial cultivars (Ciça F1, Choryoku F1, Kokushi Onaga F1, Ajimurasaki F1, Ajishirakawa F1 and Florida Market) were evaluated, coming to a total of 24 treatments arranged in randomized block design with four replications and four plants per experimental plot.
Quantitative data were initially submitted to univariate analysis of variance (p<0.01) and from the means and residual variance and covariance matrix was obtained the genetic dissimilarity matrix based on the generalized Mahalanobis distance (D 2 ). The genotype clustering was obtained by the method of ascending hierarchical classification algorithm UPGMA (Unweighted Pair-Grouped Method Average) and by the Tocher's optimization method.
The relative importance of traits in the prediction of genetic diversity was also studied through the participation of D 2 components, related to each trait in the total dissimilarity observed, and the diversity between genotypes was estimated by Mahalanobis distance. (Singh, 1981).
To test the efficiency of the hierarchical clustering method, we estimated the cophenetic correlation coefficient, obtained with 1,000 simulations, analyzed by the "t" test. The cutoff point (Cp) of the dendrogram formed by the UPGMA method was defined as proposed by Mojema (1977), following the formula Cp = m + ksd, where m = the mean distance values of the fusion levels corresponding to the stadiums; k = 1.25 (Milligan & Cooper, 1985); sd = standard deviation.
In the morphological description of the genotypes of group 1, for qualitative traits (Table 1), considerable levels of phenotypic variability were observed only for fruit color at commercial maturity, with a predominance of dark purple, followed by grayish purple, breeding programs.
Dissimilarities (D 2 ) between genotypes ranged from 1.07 to 728.53, with an average of 133.47. The largest distances were recorded between CNPH 135 and Ajishirakawa F1 genotypes. On the other hand, genotypes CNPH 47 and Florida Market were the least genetically distant (Figure 1). Thus, crossings between the most divergent groups are indicated for formation of segregating populations and with greater genetic variability for the analyzed traits.
The dendogram obtained by UPGMA hierarchical method showed the formation of three groups, considering a significant cut of 44.32% (Mojena, 1977). Group 1 was composed of most genotypes, approximately 84% ( Figure  1). Among quantitative traits, those

RESULTS AND DISCUSSION
The micrometeorological data obtained during the experiment period showed that the maximum air temperature in the greenhouse ranged between 29.8 and 41.4°C and the minimum temperature between 18.6 and 23.7°C. The average temperature ranged between 23.7 and 28.5°C. Thus, the environment was classified as high temperature for eggplant cultivation. Relative humidity ranged from 83.7 to 95.4%.
Significant differences were verified by F test (p<0.01) between genotypes for all analyzed traits (Table 1). This result refers to the existence of phenotypic variability between genotypes, and it is necessary to identify the superior genotypes to be crossed in eggplant   (Table 1). No non-commercial genotype showed considerable similarity with these commercial cultivars. G r o u p 3 i n c l u d e d o n l y t h e Ajmurasaki F1 genotype with the second longest fruit length among the evaluated genotypes (28.35 cm), smallest fruit width (2.83 cm) and highest fruit length/width ratio (9.96). (Table 1), similar to those reported by Valadares et al. (2019b). The fruits showed uniform purple coloration, snake-shaped curvature and no thorns in the fruit cup (Table 1).
Grouping of genotypes by the Tocher method was partially similar to the UPGMA method when grouping purple and green. However, the fruits showed color distribution at commercial maturation predominantly uniform with no curvature and no thorns in the fruit's cup (Table 1). This distribution indicates that, in relation to the evaluated traits (quantitative and qualitative), most genotypes presented high levels of similarity, including the commercial cultivars Ciça F1 and Florida Market, contemplated in this group 1.
According to Guedes et al. (2013), individuals are grouped in pairs, using arithmetic means of dissimilarity, and the dendrogram prioritizes genotypes with greater similarity. This explains why the Kokushi Onaga F1, Ajishirakawa F1 and Choryoku F1 genotypes formed group 2 and the Ajmurasaki F1 genotype alone group 3, consisting of fruits longer than 23.64 cm, fruit width less than 4.41 cm and length/width ratio of the fruit greater than 6.38 (Table 1). Averages for fruit length in group 2 were between 23.65 (Ajishirakawa F1) and 30.01 cm (Choryoku F1) and for fruit width between 3.20 (Ajishirakawa F1) to 4.41 cm (Kokushi Onaga F1). For length/width ratio of the fruit, the variation ranged from 1.48 (CNPH 668) to 4.91 (CNPH 84) (Table 1). For fruit color, Ajishirakawa F1 genotype presented white, Choryoku F1 green and Kokushi Onaga F1, black fruits. However, predominantly of uniform distribution and without any thorn in the fruit's cup. About fruit curvature, Ajishirakawa F1 and Choryoku F1 that fruit length, fruit width and fruit length/width ratio presented the highest percentage of contribution to divergence among the 24 evaluated genotypes, explaining 90.72% of the total genetic dissimilarity (Table 3).
High contribution of fruit length to eggplant divergence has been reported by Babu & Patil (2004) and Mehta et al. (2004), while average fruit weight and number of fruits per plant traits have lower contributions as reported by Prabakaran et al. (2015). Bashar et al. (2016) also cited contributions of length and width of fruit traits in the genetic divergence of eggplant. we observed that genotype clustering was predominantly influenced by fruit length, fruit width and fruit length/width ratio, showing greater variability for  consequently increase genetic variability (Abreu et al., 2004). Disagreements occurred in the formation of groups 3 (CNPH 84), 5 (CNPH 668) and 6 (CNPH 135) by Tocher's method. The association of clustering techniques provides a more efficient support for determination of divergence, since Tocher discriminates each group and UPGMA discriminates each genotype and can more safely infer the use of parents in breeding programs (Bertan et al., 2006).
The relative importance of the a n a l y z e d t r a i t s i n t h e g e n e t i c dissimilarity between genotypes was detected by Singh's method (1981). This method considers that the most important characteristics express greater variability. In this respect, we found among the most divergent genotypes (Table 2). Similarity between the different clustering techniques can be seen from the fact that genotypes belonging to Tocher's group 1 were mostly the same ones from the UPGMA grouping, around 71% of the genotypes, including Ciça F1 and Florida Market.
There was also agreement in the formation of group 2 which included genotypes Kokushi Onaga F1, Ajishirakawa F1 and Choryoku F1 and the formation of group 4 composed only by genotype Ajmurasaki F1. Agreement between multivariate techniques is important in the study of genetic divergence, as it allows the recommendation of crossing between the most divergent parents possible, in order to broaden the genetic base and these traits (Table 3).
According to Rohlf (2000), the adjustment of cophenetic correlation coefficient is considered good when values are equal to or higher than (r) 0.70. In this case, the greater the (r) the smaller the distortion of the cluster, presenting a good fit between the matrix and the formed dendrogram (Cruz et al., 2012).
Eggplant genotypes, under high temperatures, showed significant genetic divergence for all evaluated traits. Tocher's optimization methods and the hierarchical UPGMA agreed in 87.50% of genotypes clustering. The traits that the most contributed to divergence were fruit length, fruit width and fruit length/ width ratio. The cophenetic correlation coefficient (r) was 0.79. Most genotypes showed genetic similarity with Ciça F1 and Florida Market cultivars.