Genetic divergence among sunflower genotypes based on morphoagronomic traits in Parana State

1 Universidade Estadual de Maringá, Departamento de Ciências Agronômicas, Avenida Colombo, no 5790 CEP: 87020-900, Maringa, PR, Brasil. 2 Programa de Pós-Graduação em Agronomia – UNICENTRO, CEDETEG, Rua Simeão Camargo Varela de Sá, 03 – CEP: 85040-080, Guarapuava, PR, Brasil. 3 Universidade de Estado do Mato Grosso, Faculdade de Ciências Agro-Ambientais, Av. São João, s/no, CEP 78200-000 Cáceres, MT, Brasil.


INTRODUCTION
Cultivation of sunflower (Helianthus annuus L.) has proved to be a cost-effective option in rotations systems with other grain crops, and it is arousing the interest of farmers, agriculture professionals and companies, due to the possibility of using its oil as raw material for manufacturing biodiesel (Castro and Farias, 2005;Backes et al., 2008).In order to ensure the crop development, researches in genetic and breeding are being held for obtaining and evaluation of genotypes contemplating important aspects in the production process (Messetti and Padovani, 2004).According to Oliveira et al. (2005), efforts are needed for the *Corresponding author.E-mail: jppoletine@uem.br.development of genotypes that presents, in the same genetic material, high oil content, early cycle, reduced height, resistance to biotic and abiotic factors, besides high grain yield.
The incorporation of these characteristics of interest depends on the existence of genetic variability in germplasm available for the crop breeding programs.From this variability, it is possible to implement the selection process for the most several characteristics, searching the development of lines for hybrids constitution or to obtaining varieties of open pollination (Amorim et al., 2007).According to Barelli (2004), studies on genetic divergence have great importance in crop breeding programs since reporting parameters for identifying parents that, when hybridized, offer greater heterosis effect in progeny and a higher probability of superior genotypes recovery in segregating generations, as well as facilitate the knowledge of population genetic basis.Several works that evaluated the genetic divergence in sunflower crop were conducted by using morphoagronomic characters (Subrahmanyam et al., 2003;Messetti and Padovani, 2004;Amorim et al., 2007;Mohan and Seetharam, 2005;Arshad et al., 2007).Evaluation of genetic divergence is performed through methods based on agronomic, morphology and molecular characteristics.In the case of quantitative variables this variability may be accessed by using dissimilarity measures, outstanding Generalized Mahalanobis Distance (D 2 ii'), that considers variances and residual covariance's existing between the measured characteristics (Cruz and Carneiro, 2003).Visualization and interpretation of distances may be facilitated by the use of a clustering method and/or graphical dispersion.Clustering methods objectify to separate a group of original observations on different subgroups in order to obtain homogeneity within and heterogeneity between subgroups.Among these methods, optimization and hierarchical ones are employed on a large scale by plant breeders (Bertan et al., 2006).Genetic divergence estimation between different sunflower genotypes has been studied, aiming to develop parents for hybrids constitution or even the formation of new segregating populations, from the intercross of divergent genotypes with complementary agronomic characteristics (Amorim et al., 2007).
In this way, the work aimed to characterize the genetic variability available in different sunflower hybrids, promising in biodiesel production, and identify which agronomic characters, contribute significantly to this divergence.

MATERIAL AND METHODS
The experiment was conducted at the Regional Campus of  (Embrapa, 2006), with sandy texture.Sixteen sunflower genotypes (hybrids) from the National Assessment of Sunflower Genotypes belonging to Embrapa -Soybean Research National Center (Ending Experiment -First Year) were evaluated (Table 1), in an experimental design conducted in randomized complete blocks, with four repetitions.Each genotype was seeded in a plot constituted of four lines of 6.0 m long, spaced 0.70 m.The distance between plants was 0.30 m, totaling 21 hills per line.Each hill contained three and seeding density ranged between 40,000 to 45,000 plants per hectare.Roughing was conducted, seven days after emergency, resting only 21 plants per line.
In addition to the recommended fertilization (40 to 60 kg per hectare of N, 40 to 80 kg per hectare of P2O5 and 40 to 80 kg per hectare of K2O, it was applied one fertilizer containing boron (B), 2.0 kg per hectare in the soil, mixed with coverage nitrogen fertilization, 25 days after plants emergence.Control of weeds, insects and diseases was developed according to crop needs, by using the recommended chemical products for the crop.The harvest of each plot was performed manually.
Two central lines were harvested (useful area), eliminating 0.50 m from each edge.Chapters were covered with TNT protection sacks to prevent birds attack.
The following characteristics were analyzed: ( In the analysis with canonical variables it was evaluated the progenitor similarity through a graphical dispersion, while agglomerative methods are dependent on the dissimilarity measures estimated through Mahalanobis Generalized Distance (D 2 ii') (Cruz and Regazzi, 1994).Genotypes clustering was performed by using Optimization Method proposed by Tocher, cited by Rao (1952), according to Cruz andCarneiro (2003), andOliveira et al. (1998), where individuals belonging to the same group are more homogeneous than individuals of different groups, and    Hierarchical Method "UPGMA" with genotypes clustered by a process that repeats for several levels until it is established the dendrogram, where demarcations may be established by a visual examination of this tree in which occurs the assessment of points with high change level, taking them as delimiters of genotypes number in order to determine a group (Cruz and Regazzi, 1994).

Genotype
Both variance analysis and multivariate analysis were performed by using Gene's computational program (Cruz, 2006).

RESULTS AND DISCUSSION
By variance analysis (Table 2) there were found significant differences (1% probability level), for GRAIN, TC and CC characteristics.STD and OIL (%) showed significant differences at 5% probability level, by F test, demonstrating the existence of variability among genotypes and suggesting that evaluated characters are important in characterizing genetic divergence, with the exception for WTA and PH parameters, that showed no significant differences.Variant coefficient ranged from 5.67 to 19.70%, revealing appropriate experimental precision, close to the ones found by Amorim et al. (2007) and Vogt et al. (2010).
Table 3 shows mean values for the seven evaluated characteristics, with Scott and Knott clustering at 5% probability level.For GRAIN parameter (grain yield per hectare), it was observed the constitution of five distinct groups, with mean yield around 2632.94 kg ha -1 , with SYN 045 hybrid the most productive, followed by the second group, that contemplated M 734 (T), HLA 05-62, HLA 11-26, QC 6730 and SYN 039A hybrids. Balbinot Jr. et al. (2009), in genotypes evaluation experiment in 2007/2008 agricultural year, for the same seeded period, in Santa Catarina State, obtained mean yield inferior to this study, as much as cultivars and hybrids.This happened because of water deficiency during grain filling, showing in chapter size, achenes with reduced mass.It is worth mentioning that during the grain filling period (December to January) in this present study, accumulated rainfall exceeds 240 mm, what contributed to higher grain yield.Matter et al. (2009), in South region of Brazil, also found maximum grain yield around 2485 kg ha -1 , studying similar genotypes.
Observing the results obtained by Embrapa (2011), it is possible to verify that SYN 045 hybrid, also presented high grain yield in Coxilha and Rio Pardo Counties, Rio Grande do Sul State and Londrina County, Parana State.For the characteristics weight of 1000 achene's, final stand and plants height, it was observed the formation of only one group, by the statistical method used, indicating that there is no genetic variability for 16 sunflowers hybrids, in relation to these parameters.Vogt et al. (2010), in sunflower cultivars competition experiments at North region of Santa Catarina State, observed similar results to this present study, for mean values of chapter size, chapter curvature and weight of 1000 achene's.To WTA characteristic, hybrids showed mean of 55.15 g, less than the values obtained by Backes et al. (2008) andBalbinot Jr. et al. (2009), but higher than the results from Amorim et al. (2008).
Plants height ranged from 174.8 cm (Helium 358 (T) hybrid) and 224.5 cm (QC 6730 hybrid), corroborating the extent of results obtained by Castiglioni et al. (1994) and Embrapa (2011), but lower than data published by Amorim et al. (2008).The mean chapter size reached 23.86 cm, agreeing with Castro et al. (1996) and Rossi (1998) who cited that this characteristic ranges from 6 to 50 cm depending on the genotype, with higher values achieved by hybrid genotypes and confirming the results obtained by Amorim et al. (2008).
Chapter curvature showed mean value around 4.31, with evaluations conducted according to Knowles (1978) methodology, closer to the means obtained by Amorim et al. (2007) and Vogt et al. (2010).Coimbra et al. (2009), studying sunflower hybrids, genetically similar to the ones of this present study, in Palmas County, Tocantins State, observed mean values for plants height (115.77cm), grain yield (1280.8 and 1280,8 kg ha -1 ), weight of 1000 achene's (48.5 g), chapter size (14 cm) and oil content (47.8%), lower than those obtained for Umuarama County, in this analysis.It should be noted that in the experiment conducted in Tocantins State, sowing occurred in February, in a place with higher temperature and means rainfalls throughout the cycle around 290 mm, with lower water demand of the experiment in this study area, what may cause decrease in grain yield and oil content (Acosta, 2009).
When the evaluation refers to oil content, Scott-Knott test separated genotypes in two groups, with mean oil content around 50.3%, considered ideal for the crop, agreeing with the results obtained by Coimbra et al. (2009), whose study revealed mean levels around 47.8%.According to Embrapa (2011) or crosses between SYN 045 x SULFOSOL, SYN 045 x HLA 44-49, SYN 045 x SYN 034A, SYN 045 x V70153 and SYN 045 x PARAÍSO 65 genotypes, which would express the greatest heterotical potential and better use in crop breeding programs.
When calculating the relation of distance between a genotype in the presence of the others, it was observed that SYN 045 genotype composed five combinations with high dissimilarity values with magnitude of D 2 ii ' around 134.54; 131.61; 111.40; 108.03 and 100.48% to   Group II genotype is caracterized by higher grain yield, lower size chapter, higher chapter curvature and lower oil content.
It is important to highlight that methodologies used for genetic divergence analysis (Tocher Optimization and Hierarchical Method "UPGMA") are based on the same Dissimilarity Matrix (Table 4).However, the calculation used for genotypes grouping is originated from different analysis, considering UPGMA method as more discerning, showing the formation of subgroups (Figure 1).Thus, it is common in studies of genetic divergence, the presentation of both methodologies (Cruz, 1990;Amorim et al., 2007;Vogt et al., 2010), with comparative purposes.This way, it was possible to observe that, being characterized as more detailed, genetic divergence obtained through Hierarchical method "UPGMA" pointed to the formation of subgroups, agreeing with Tocher Optimization Method, positioning SYN 045 genotype ( 16) disconnected.This situation was already expected because of its character highly productive, confirmed in Table 3 (agronomic characteristics analysis) and by canonic variables through scores of graphical dispersion (Figure 2).
Cofenetic correlation coefficient (CCC), applied to   6) it is possible to verify that the first two variables explained 82.07% of total variation (68.03% for first and 14.04% for the second one), enabling the transposition of genetic divergence of p-dimensional space (p = 7, in this case) to the bi-dimensional with negligible degree of distortion, caused by the distances between the genotypes.Graphical analysis, in comparison studies of similaritybetween cultivars must be considered when it is possible to summarize in a few variables more than 80% of the total available variation.For Bock (1975), if the first canonical variables accumulate 70% or more of total available variation between individuals assessed, the descriptors may be substituted for them.The results obtained in this study allow to demonstrate with trust the graphical dispersion, in relation to the first two canonical variables in bi-dimensional space (Figure 2), where it is possible to observe total agreement of graphical dispersion with UPGMA clustering method, and partially agreeing with the methodology proposed by Tocher.
Knowledge of relative importance of the characteristics for genetic divergence allows discarding characteristics with little contribution, thereby reducing manpower, time and costs spent on experimentation (Cruz, 1990).The criterion proposed by Singh (1981), based on Mahalanobis Generalized Distance ( 2 ' ii D ) evidenced that GRAIN and SC characteristics (Table 7) were the ones that most contributed to divergence characterization, with 62.73 and 13.47%, respectively.CC (9.01%), OIL (4.97%), STD (4.35%), PH (3.55%) and WTA (1.88%) characteristics, showed minor contribution being WTA the suggested variable for disposition.Through the ratio of of CVg and CVe greater than unity, it is possible to use with ease the variables GRAIN and CS in selection, since also presented high heritability values.
Results obtained by Amorim et al. (2007), aiming to estimate the relative contribution of each characteristic to the expression of genetic divergence, indicated that flowering initial (13.10%),flowering days (37.10%), height of chapter insertion (18.55%) and number of leaves (9.10%) characteristics were the ones that most contributed to total divergence among 15 sunflower genotypes (78.05%).The authors still commented that Alvarez et al. (1996) verified that flowering initial and number of leaves were also important in discriminating genetic divergence among sunflowers populations.Such characters were not evaluated in this work, but due to its importance, it will be subject to further studies.
It is observed through the ratio between Genetic Variation Coefficient (CVg) and Experimental Variation Coefficient (CVe), value greater than the unit for GRAIN and CS characteristics, inferring by the ease on selection for such characters.Besides it, the high heritabilities (96.65%) and (82.70%) respectively, also contribute to selection procedures.According to Amorim and Souza (2005), the lines obtaining from hybrids is a viable alternative, since the behavior of the genotypes has been studied in distinct environments and it is still possible to find satisfactory ratio of favourable locos already fixed.

Conclusion
Although restricted, there is genetic variability among 16 sunflower hybrids for the evaluated agronomic characteristics, except for weight of 1000 achene's and plant height.Grain yield and chapter size characteristics contributed significantly in genetic divergence observed among hybrids.It is possible to identify divergent genetic materials for obtaining lines and/or formation of new populations, aiming biodiesel production, in conditions of Parana State North western region.
,54), being allocated in Groups I and II, respectively.Crosses more similar revealed by Mahalanobis Generalized Distance were obtained between HLA 44-49 x SULFOSOL, HLA 05-62 x HLA 11-26, M 734 (T) x HLA 11-26 and HLA 44-63 x V60415 genotypes, belonging to the same group (I) by Tocher clustering, concurring with Mahalanobis methodology.The dendogram (Figure 1) generated by Unweighted Pair Group Method with Arithmetic Mean (UPGMA) was constituted by three groups.

Figure 2 .
Figure 2. Scores of graphical dispersion, in relation to two axes representing the first two Canonical Variables (CV1 and CV2), obtained from seven characteristics evaluated in 16 sunflowers cultivars (Umuarama -PR, 2011).

Table 1 .
Sunflowers genotypes used in the experiment, obtaining company and origin country (Umuarama, PR/2011).