AGRONOMIC CHARACTERIZATION AND GENETIC PARAMETER ESTIMATION IN YELLOW PASSION FRUIT CARACTERIZAÇÃO AGRONÔMICA E ESTIMATIVA DE PARÂMETROS GENÉTICOS EM MARACUJAZEIRO AZEDO

The selection of yellow passion fruit (Passiflora edulis Sims) genotypes with high yield and fruit quality is essential for the development of passion fruit crop in the country. Therefore, the objective of this study was to evaluate the agronomic performance and estimate the genetic parameters of 32 yellow passion fruit genotypes cultivated in the Federal District, Brazil. The experiment consisted of randomized block design with 32 treatments, eight plants per plot, and four replications. Fruits were classified based on their equatorial diameter. Yield, number of fruits per hectare, and fruit weight were evaluated. Genotypes MAR20#23 and UnB-P7 presented the highest total yield, and MAR20#23 also showed the greatest total number of fruits per hectare. BRS GA1 and MAR20#23 had the best performance for industrial purposes dues to the fruits of smaller diameter classes. UnB-P7, AR-01, and MSC were more indicated for in natura consumption owing to the fruits of greater diameter classes. The high magnitude estimates for heritability and genetic variation coefficients indicate the possibility of greater genetic gains with direct selection for yield and number of fruits of 1C diameter class. Significant phenotypic correlations were observed, indicating the possibility of indirect selection for number of fruits, fruit weight, and yield.


INTRODUCTION
Brazil stands out as the world's largest producer and consumer of passion fruit. In 2016, fruit production reached 703,489 tons in an area of 49,889 hectares (IBGE, 2016). Yellow passion fruit (Passiflora edulis Sims), also known as sour passion fruit, represents approximately 95% of the national production (COSTA et al., 2008). The Brazilian passion fruit mean yield (14.10 t ha -1 ) (IBGE, 2016) is considered as low when compared with the productive potential of the species, which can reach more than 40 t ha -1 (FREITAS et al., 2011;NEVES et al., 2013). The cultivation of inadequate varieties (JUNQUEIRA et al., 1999), the low production technology usage (MELO et al., 2001), and the lack of homogeneous and productive materials (MELETTI et al., 2000) are limiting factors to the increase of fruit quality and orchards yield.
The development of genotypes with high yield, uniformity, and fruit quality is important for breeding programs (FALEIRO et al., 2006) since they increase the species production potential (NEVES et al., 2013). In addition, these materials may increase the producers' income and simplify some stages of the productive process, such as post-harvest fruit classification (MELETTI et al., 2000).
Fruit quality is measured both by its internal and external characteristics. The former is related to flavor (acidity and sugars content) and juice content (yield). The latter is associated with appearance and fruit standardization parameters, which influence the consumer's choice. Fruits with better external appearance are usually intended for the fresh fruit market, while the others are destined for the industry (BALBINO, 2005). Therefore, the objective of this study was to evaluate the agronomic performance of 32 yellow passion fruit genotypes cultivated in the Federal District, Brazil, and to estimate some essential genetic parameters for the definition of breeding strategies.

MATERIAL AND METHODS
The experiment was conducted at Água Limpa Farm, belonging to University of Brasilia (UnB) (16 o S and 48 o W, 1,100 m asl), located in Brasilia, DF, Brazil. It consisted of a randomized , where is the estimator of phenotypic covariance between two traits X and Y; is the estimator of the phenotypic variance of trait X; and is the estimator of the phenotypic variance of trait Y.

RESULTS AND DISCUSSION
Significant differences were observed between genotypes for all traits, except for fruit weight of diameter classes 2A and 3A. Studies carried out with yellow passion fruit progenies have demonstrated high genetic variability for some traits, such as number of fruits per plant, and fruit weight and length (FREITAS et al., 2012;KRAUSE et al., 2012). This variability evidences the breeding potential of genotypes selection (OLIVEIRA et al., 2008).
The total yield varied from 4,055.19 (EC-3-0) to 15,474.39 kg ha -1 (MAR20#23), with a mean of 9,543.88 kg ha -1 for all genotypes (Table 1). For some of the genotypes studied, the total yield was higher than the national mean (IBGE, 2015). Studies on the agronomic traits of different yellow passion fruit genotypes cultivated in the Federal District have demonstrated distinct behavior for EC-3-0. According to these data, genotype EC-3-0 presented superior performance, with a total yield of 15,460 (ABREU et al., 2009) and 26,480 kg ha -1 (JUNQUEIRA et al., 2003). Coimbra et al. (2012), Maia et al. (2009), andJunqueira et al. (2003) observed a superior performance for EC-RAM in relation to the other genotypes, with an estimated total yield of 43,287; 13,968; and 32,880 kg ha -1 , respectively. However, in the present study, EC-RAM exhibited a total yield of only 6,350.40 kg ha -1 .
The total number of fruits per hectare ranged from 31,063 (UnB-P5) to 119,715 (MAR20#23). In general, the most productive genotypes also presented the greatest number of fruits. Although they did not significantly differ, a considerable numerical variation of 52,000 fruits between AR-01 and EC-RAM was recorded. These fruits, in practical terms, could be commercialized (Table 1). MAR20#23 can produce 342,847 fruits per hectare in the edaphoclimatic conditions of the Federal District (COIMBRA et al., 2012), confirming the superior performance of the genotype in this region. Fruits of diameter classes 1C and 1B are of great interest to the industry for being small and usually rejected by the fresh fruit market. Fruits of greater diameter classes (1A, 2A, and 3A) are intended for the in natura market (COIMBRA et al., 2012). The highest yield and the greatest number of fruits of 1C diameter class were observed in BRS GA1, with 3,343.41 kg ha -1 and 53,939 fruits, respectively. MSC presented the lowest yield and number of fruits of 1C diameter class, with 238.14 kg ha -1 and 4,455 fruits. For 1B diameter class, MAR20#23 presented the highest yield (9,019.51 kg ha -1 ) and the greatest number of fruits (62,875). MSC showed the lowest yield (2,295.98 kg ha -1 ) and UnB-P5 showed the lowest number of fruits (16,192). For fruits classified as 1A, UnB-P7 presented the highest Biosci. J., Uberlândia, v. 34, supplement 1, p. 58-70, Dec. 2018 yield (4.156.60 kg ha -1 ) and number of fruits (19,251). In contrast, EC-3-0 showed the lowest yield (643,60 kg ha -1 ), and MAR20#29 exhibited the lowest number of fruits (3,480) ( Table 2).
Low yields were observed for larger fruits (2A and 3A) when compared with the other diameter classes. For 2A diameter class, AR-01 presented the greatest fruit yield (639.99 kg ha -1 ) and the greatest number of fruits (2449), differing only from genotype MAR20#29, which registered 2.86 kg ha -1 and 13 fruits ( Table 2). This result corroborates other studies that reported the superior performance of AR-01 in relation to fruit yield and number of fruits in the diameter classes 1A and 2A (ABREU et al., 2009;MAIA et al., 2009;COIMBRA et al., 2012). For the fruits of 3A diameter class, MSC presented the best performance, with 115.94 kg ha -1 and 409 fruits ( Table 2).
Greater fruit weight is fundamental to the fresh fruit market. Larger fruits have better commercial classification and, consequently, better prices. Neves et al. (2013) stated that fruits with fruit weight greater than 180 g present good commercial value for the fresh fruit market. The total weight per fruit ranged between 84.56 g (Roxo Australiano) and 155.25 g (MSC), with a mean weight of 121.58 g for all genotypes evaluated (Table 1). Coimbra et al. (2012) reported similar results for genotypes AR-01, AR-02, MAR20#03, FP-01, RC-3, and EC-RAM, and did not detect differences between genotypes. However, the fruit weight range obtained in this study, as well as that reported by Coimbra et al. (2012), was lower than the variations recorded by Campos et al. (2007) (191 and 228 g fruit -1 ) and by Vale et al. (2013) (207 and 286 g fruit -1 ). Considering the diameter classes 1C and 1B, genotypes UnB-P7 (71.25 g) and MAR20#15 (155.25 g) presented the highest fruit weight, respectively, while genotypes EC-RAM (38.06 g) and Roxo Australiano (99.00 g) showed the lowest values, respectively. UnB-P5 obtained the greatest fruit weight in the 1A diameter class (231.56 g), differing only from EC-3-0 (174.56 g). No significant differences were recorded between fruit weight in the diameter classes 2A and 3A for the different genotypes ( Table 2).
The contrasts observed among the different studies may be due to the genetic nature of the analyzed materials since they are originated from open pollination. Additionally, different water and nutritional management, distinct harvest times, different substrates and packages used for seedling formation, plant physiological conditions, and environmental factors favorable to the occurrence of phytopathogens may have contributed to these variations. The pollination type should also be considered since it influences fruiting. Artificial pollination, which was not performed in the present study, would have substantially increased yield, fruit weight, fruit diameter and length, and percentage of pulp; and reduced peel thickness (KRAUSE et al., 2012). Such traits would favor both the fresh fruit market and the industry. Fruits with greater pulp weight and size are preferred by the in natura market. Conversely, high pulp yield and reduced peel thickness are desirable traits for industrial purposes (FERREIRA et al., 2010).
The three variables evaluated presented heritability estimates of high magnitude (Table 3). The CV g for the variables total yield (14.70), total number of fruits (15.06), and total fruit weight (4.09) were low when compared with the results obtained by Freitas et al. (2011) (Table 3). These low values may indicate the existence of low genetic variability among the evaluated genotypes.
Heritability values ranged from 8.17 (fruit weight of 2A diameter class) to 81.70% (number of fruits of 1C diameter class) (Table 4). Considering the magnitudes of heritability estimates and the CV g , higher gains with direct selection are expected for yield and number of fruits of 1C diameter class since the environment has little influence on the expression of these traits. Oliveira et al. (2008) reported the highest estimates of coefficients of heritability (greater than 50%) for fruit length, number of fruits per plant, and fruit weight. Similarly, Freitas et al. (2011) reported high heritability values for the number of fruits per plot (98.02%) and fruit length (82.69%).
Despite the low CV g observed for the different diameter classes, the CV g /CV e ratio was greater than 1 for number of fruits of 1C diameter class (1.06) and total fruit yield for the same diameter class (1.02) ( Table 4). These values indicate a favorable selection condition since the genetic variance surpasses the environmental variance. These variables also presented the highest heritability estimates, contributing to the values of CV g /CV e ratio. For the other variables, CV g /CV e ratio was lower than 1, indicating that the use of simple breeding methods, such as mass selection, will not provide significant gains during the selection process. Breeding methods based on family performance are more appropriate Biosci. J., Uberlândia, v. 34, supplement 1, p. 58-70, Dec. 2018 than those based on individual performance (FREITAS et al., 2015). Seventy significant phenotypic correlations were observed, with valuesvarying between 0.34 and 0.99. In 60% of the significant correlations, estimates values were equal to or greater than 0.6, indicating a strong correlation and the possibility of indirect selection for these traits. The presence of significance for low correlation values has been detected by other authors (VASCONCELOS et al., 1998;KUREK et al., 2002), and may be associated with the high degrees of freedom included in the t-test (VASCONCELOS et al., 1998) (Table 5). In this study, the increase in the total number of fruits increased the total yield (r f = 0.96), as also reported in other studies (OLIVEIRA et al., 2011;NEVES et al., 2013). Similarly, Pimentel et al. (2008) recorded a strong phenotypic correlation between number of fruits and yield per plant. Strong phenotypic correlations were observed between number of fruits and fruit yield for each diameter class: 1C (r f = 0.98), 1B (r f = 0.97), 1A (r f = 0.99), 2A (r f = 0.99), 3A (r f = 0.98). For fruits of 3A diameter class, a strong correlation was reported between number of fruits and fruit weight (r f = 0.92), and between fruit yield and fruit weight (r f = 0.94) ( Table 5).  Table 3. Estimates of mean phenotypic (V p ), genotypic (V g ), and environmental variances (V e ), mean broad sense heritability (h 2 ), coefficient of genetic variation (CV g ), and coefficient of relative variance (CV g /CV e ) for total yield (TY; Kg ha -1 ), total number of fruits per hectare (TNF), and total fruit weight (TFW; g) in 32 yellow passion fruit (Passiflora edulis Sims) genotypes cultivated in Brasilia, DF, Brazil.
The total number of fruits was negatively correlated with the fruit weight (r f = -0.13), as also observed by Pimentel et al. (2008) and Neves et al. (2013) for genetic correlations in parents and hybrids of yellow passion fruit. Therefore, an increase in the number of fruits reduced fruits size, which is undesirable for the fresh fruit market. The total number of fruits negatively correlated with the number of fruits of 1C (r f = -0.34) and 1B (r f = -0.10) diameter classes, as well as with the yield of fruits of 1C diameter class (r f = -0.24). A negative correlation was also reported between the variables of fruits of smaller diameter class (1C and 1B) and the variables of fruits of greater diameter class (2A and 3A) ( Table 5).

CONCLUSIONS
Genotypes MAR20#23 and UnB-P7 presented the highest total yield, with values higher than the national mean. MAR20#23 also showed the greatest total number of fruits per hectare, standing out from the other genotypes.
The highest yield and the greatest number of fruits suitable for industrial purposes (diameter classes 1C and 1B) were identified in genotypes BRS GA1 and MAR20#23.
For the in natura consumption, the best performance was verified in UnB-P7, AR-01, and MSC (diameter classes 1A, 2A, and 3A, respectively).
High heritability values and CV g /CV e ratio greater than 1 for yield and number of fruits of 1C diameter class indicate a favorable selection condition.
The differences in yield and fruit quality recorded for the genotypes of this study confirm the potential for selection and future crosses that aim at increasing these traits.