Genetic divergence among accessions of Carthamus tinctorius L. by morphoagronomic traits

Safflower (Carthamus tinctorius L.) is a crop that produces oil with high content of oleic and linoleic acids, which has great quality for human consumption and industrial use, such as in biodiesel production, and cosmetics and pharmaceuticals. The objective of this study was to quantify the genetic diversity among 20 safflower genotypes by morphological characteristics. The genotypes were grouped into nine groups based on the measurements of genetic divergence obtained by means of Mahalanobis distance (D 2 ) and in the grouping by Tocher optimization method. Heritability coefficients ranged from 89 to 93% and b quotient was above 1.0 indicating that the variability in the population allows obtaining genetic gains for all traits. Grain yield is the character that most contributed to the genetic variability of the population studied. The accesses of Group V are favorable for use as parents in crosses to obtain new genotypes with high grain and oil yields, and desirable plant architecture.


INTRODUCTION
Safflower (Carthamus tinctorius L.), is a species that belongs to the Asteraceae family (Khan et al., 2009), is cultivated for many purposes.The oil extracted from its seeds has high levels of oleic and linoleic acids of excellent quality for human consumption and industrial use (Hamdan et al., 2009).The safflower flowers are used as sources of dyes and for medicinal purposes (Koutroubas et al., 2009).The achenes are used to feed birds, and the plant is used in hay production as well as for ornamentation (Emongor, 2010).The safflower pie, a byproduct of oil extraction, has about 25% protein and is used in ruminant feed (Ekin, 2005).
The crop develops well under semi-arid conditions, in a diverse range of soil types.Safflower plants have high tolerance to drought, high temperatures, windy conditions and low relative humidity (Kizil et al., 2008;Bagheri and Sam-Dailiri, 2011).They also have moderate salinity tolerance (Feizi et al., 2010) and resist a wide range of temperatures from -7 to 40°C, depending on their stage of development (Emongor, 2010).The importance of safflower as an oil crop has increased in recent years due to the increasing interest in biofuel production (Dordas and Sioulas, 2008).Demand for raw material extracted from safflower is also increasing in food, medicine, and cosmetics industries, among others.To meet this demand, an increase in *Corresponding author.E-mail: acarneiroagro@yahoo.com.br.Tel: +55 14 3811 7161.Fax: +55 14 3811 7102.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License Head diameter (cm), e TWG: Thousand grain weight (g), f GY: Grain yield (kg ha -1 ), g WDM: Dry matter weigh (kg ha -1 )t, h HI: Harvest index, i OC: Oil content (%), j OY: Oil yield (kg ha -1 ).
production is necessary, both by increasing productivity and expanding cultivated areas.Therefore, the improvement of safflower varieties already cultivated is necessary as well as obtaining new genotypes.For the improvement of existing safflower varieties, selection of several suitable parents for hybridization is an important step to obtain the desired recombination in segregating generations (Shivani et al., 2011).Diverging genotypes with complementary agronomic traits are necessary in hybrids formation or new segregating populations (Barbieri et al., 2005;Arriel et al., 2007).More diverging genitors result in greater variability in the segregating population, and larger probability of regrouping alleles in new favorable combinations (Barbieri et al., 2005).Quantification of genetic diversity that is performed by morphological and agronomic traits, considered as quantitative traits, can be accessed using dissimilarity measurements, such as: Euclidian distance and Mahalanobis distance (D 2 ).The latter takes into account the existing residual variances and covariances among the measured characteristics, when the experiment is under experimental design (Cruz and Carneiro, 2006).
The knowledge of genetic parameters such as heritability (h 2 ), genotypic coefficient variation (CV g ) and index of variation (b) is of great importance for breeders, as it guides the choice of the most appropriate method for crop improvement, maximizing gains from selection (Cruz, 2005).Therefore, the objective of this study was to quantify the genetic divergence among safflower accessions, estimate genetic parameters and identify the contribution of morphoagronomic characters in the genetic variability of the population in order to assist in the selection and development of superior safflower genotypes.

Experimental details
A field experiment was conducted at the Lageado Experimental Farm (750 m altitude, 22°49'31 "S, 48°25'37" W) from the São Paulo University State, in Botucatu, São Paulo, Brazil, from April to September, 2011.Twenty safflower genotypes were cultivated, in which 19 accessions were from North American Western Regional Plant Station (WRPS), located in Pullman, Washington in the United States, and one was a commercial cultivar from Paraguay.The experimental design was a randomized blocks with three replications.The experimental plot consisted of four 3 m rows, with 0.5 m between rows and 0.2 m between plants.Seeds were sown at a density of 10 seeds per meter and thinning was performed 30 days after sowing, remaining 5 plants m -1 , which resulted in a population of 200,000 plants ha -1 .Plants were only rainfed without use of any supplemental irrigation throughout the season.

Plant sampling and measurements
The traits were: plant height, number of branches per plant, number of heads per plant and head diameter was obtained from an average of 10 plants randomly selected and evaluated in each experimental plot.Thousand grain weights were obtained by averaging four samples of 1000 grains for each plot, with water content adjusted to 10%.Grain yield was obtained by harvesting grains from 10 plants per plot, and transforming g plant -1 to kg ha -1 , with water content adjusted to 10%.Dry matter per plant was obtained in a total weight of 10 plants, after drying in circulating air oven at 65°C for 72 h.Harvest index resulted from the ratio of the grain mass divided by the total mass of plant, which was obtained from average of 10 plants per plot.The oil content was determined by the method of Nuclear Magnetic Resonance NMR (Oxford, 1995) and the oil yield was obtained by a relationship between the oil content and seed productivity.

Statistical analysis
Analysis of variance F test (p < 0.05) was performed for all traits.The measurements of genetic diversity were obtained by Mahalanobis distance (Mahalanobis, 1936), and accessions were clustered by the Tocher optimization method.The criterion of Singh (1981) was used to quantify the relative contribution of the traits to genetic divergence.The estimates of genetic parameters environmental variation coefficient (CVe), genetic variation coefficient (CVg), b quociente (CVg/ CVe ratio) and heritability coefficient in broad sense (R 2 ) were calculated based on the expected mean squares (Vencoviscky and Barriga, 1992) and the simple correlation coefficients (r) between the traits were estimated by the Pearson method.The analyses were performed using Genes Software (Cruz, 2001).

Analysis of variance and cluster analysis
The analysis of variance shows that there were significant differences among genotypes for all traits evaluated (Table 1), showing the presence of genetic  variability among genotypes as well as the possibility of genetic gains for all the traits.Cluster analysis by the Tocher optimization method allowed the separation of 20 genotypes into nine groups (Table 2).The principle of this method is to maintain homogeneity within groups and also heterogeneity between groups.
In Group I, 6 genotypes were included while the Groups II and IV were formed by 3 genotypes, Groups III and V were composed of 2 genotypes, and the other groups were formed by only one genotype (Table 2).The formation of these groups is of great importance to the choice of the progenitors, as it allows the identification of the divergence magnitude between genotypes and assists in choosing the best hybrid combinations to be used in a breeding program.The divergence values (D 2 ) within and between groups are shown in Table 3.The minimum distance intra groups (26.990) were found in Group III, and the maximum distance (45.053) in Group V. The minimum distance inter groups (48.479) was found between the Groups II and VIII, and the maximum distance (180.452) in the Groups V and IX.
Considering that larger inter groups distance is positively related to larger genetic divergence, crosses between the genotype of the Groups V and IX are of greatest potential for the formation of new segregating populations with greater genetic variability.However, considering the low grain yield and oil in Group IX, these combinations are not beneficial.Although genetic divergence is one of the most important criteria for the choice of the progenitors, the hybrid combinations should involve parental both divergent and of high average performance (Abreu et al., 1999).Thus, the selection of the most divergent genotypes is neither always the most convenient for the breeding program, depending on the characteristics desired to improve.
Table 4 shows the average of the traits studied for the nine groups formed.From the average of the traits, it is possible to perform the selection based on both divergence and performance of genotypes.The average for the Group I is closer to the general population average for the majority of the traits (Table 4), which explains the larger number of genotypes in this group (30 %).Genotypes included in the Group V are of better performance for grain yield (1833 kg ha -1 ) and oil yield (502 kg ha -1 ) (Table 4), which are the traits of most interest to the of safflower breeding.Study conducted by Shivani et al. (2013) indicated that the grain yield for 12 groups formed from 90 safflower genotypes ranged from 351.67 to 1608.50 kg ha -1 . Beyyavas et al. (2011), evaluating 26 safflower genotypes, observed that oil yield GY: Grain yield (kg ha -1 ), g WDM: dry matter weight (kg ha -1 ), h HI: Harvest index, i OC: Oil content (%), j OY: Oil yield (kg ha -1 ). .The average for the number of head per plant of the Group V, the largest group in this study (Table 4), is similar to the average of superior genotypes for that trait evaluated by Shivani et al. (2011) and Beyyavas et al. (2011).
The oil content ranged from 20.7 to 30.5% (Table 4).Shivani et al. (2011) observed that the oil content ranged from 24.4 to 28.21% among the genotypes evaluated.In a study performed by Beyyavas et al. (2011) the oil content reached 34.8% for the more productive genotype.By considering the existence of genotypes in other populations with better performance for grain yield, oil yield and traits related to productivity, such as number of heads per plant and oil content, the introduction of new genotypes superior for these characteristics through new hybrid combinations can provide additional gains for grain and oil yields to the genotypes from the studied population.The mean values of the traits show that genotypes of Group V have potential to be used as progenitors for obtaining new genotypes with high grain and oil yields as well as desirable plant architecture, through indirect selection for traits such as number of branches per plant and plant height, which have influence on important aspects for cultivation, such as mechanical harvesting, resistance to lodging and plant density.
Considering the high grain and oil yields of Group V and that the distance intra group for this group is the largest among the groups (Table 3) the cross between the genotypes of Group V is an option of combination to be considered to maintain high productivity of grains and oil in the descendants and still obtain variability for the other traits.The analysis of the relative importance of the traits shows that the highest percent of genetic variability among genotypes is attributed to grain yield (Table 5), indicating that this was the trait of greatest contribution to the genetic variability of the population studied.The number of heads per plant and dry matter weight were the traits of lower contribution in genetic variability of the  population and in the formation of groups (Table 5).

Estimates of genetic parameters
The estimates of the experimental variation coefficients (CVE) ranging from 5.36% for MMG to 22.26% for PMS(Table 5) indicate that the accuracy of the experiments was satisfactory (Table 6) according to the classification Pimentel-Gomes and Garcia (2002).Estimates of genetic variation coefficients (CVg) indicate the presence of genetic variability for all traits, which allows the selection for plant breeding.The traits oil yield, grain yield, dry matter weight and number of heads presented higher values of CVg suggesting ease for selection and genetic gains expected with these traits in relation to traits such as head diameter and oil content with lower values of CVg (Table 6).
The b quotient values above 1.0 (Table 6) indicate that genetic variation overcomes environmental variation, providing thereby a favorable condition for the selection to be performed with high efficiency and less interference from the environment (Vencovsky and Barriga, 1992).Estimates of the heritability coefficients of high magnitude (R 2 > 0.7) (Table 6) show that most of the phenotypic variation is attributed to genetic causes, and also there are wide possibilities of genetic gains for all the traits.This can be explained due to the fact that genotypes have distinct origins (USDA, 2015) and have not been previously subjected to advanced stages of selection.

Correlation analysis between traits
Grain yield was positively and significantly correlated with the dry matter weight, number of branches per plant and number of heads, suggesting that a selection to improve these traits may contribute to increase grain yield (Table 7).The oil yield had positive and significant correlation with grain yield, dry matter weight, number of branches per plant, number of heads and oil content (Table 7).Beyyavas et al. (2011), observed a positive and significant correlation between the oil yield and seed yield, number of heads per plant, seed weight and oil content.
In addition to the correlation between themselves, both the grain yield and oil yield are correlated with the dry matter weight per plant, number of branches and number of heads per plant.Therefore, these characters should be considered in selection of progenitors for use in safflower breeding programs in order to obtain genotypes with higher productivity.

Conclusion
The genetic divergence between the safflower access and the high heritability coefficients and genetic variation enable obtaining genetic gains for all traits evaluated.Grain yield was the trait that most contributed to the genetic variability of the population studied.The genotypes of Group V are promising to be used as progenitors in crosses to obtain new genotypes with high grain and oil yields as well as desirable plant architecture.
0.05, by the T test.a TWG: Thousand grain weight, b GY: Grain yield, c WDM: Dry matter weight, d BP: Branches per plant, e HP: Head per plant, f HD: Head diameter, g PH: Plant height, h HI: Harvest index, i OC: Oil content, j OY: Oil yield.

of variation GL Mean square PH a BP b HP c HD d TWG e GY f WDM g HI h OC% i OY j
* p < 0.05.a PH: Plant height (cm), b BP: Number of branches per plant, c HP: Head per plant, d HD:

Table 2 .
Grouping of 20 genotypes of C. tinctorius L. by Tocher method as a function of the generalized Mahalanobis distance (D 2 ).

Table 4 .
Cluster means of twenty C. tinctorius L. genotypes.Plant height (cm), b BP: Number of branches per plant, c HP: Head per plant, d HD: Head diameter (cm), e TWG: Thousand grain weight (g), f a PH:

Table 5 .
Relative importance of 10 morphoagronomic traits in genetic diversity among 20 genotypes of C. tinctorius L.

Table 6 .
Estimates of genetic variation coefficients for 20 genotypes of C. tinctorius L.

Table 7 .
Estimates of simple correlations (r) between 10 traits evaluated in 20 genotypes of C. tinctorius L.