Evaluation of the reaction of sunflower inbred lines and their F1 hybrids to drought conditions using various stress tolerance indices

In this study, 21 genotypes of sunflower (Helianthus annuus L.) derived from a half diallel cross between six inbred lines were evaluated in both water-stressed and well-watered environments. In each environment, the genotypes were evaluated using a randomized complete block design with three replications. From the grain yield data, drought tolerance indices comprising of stability tolerance index (STI), mean productivity (MP), geometric mean productivity (GMP), harmonic mean (HM), stress susceptibility index (SSI), tolerance index (TOL), yield index (YI) and yield stability index (YSI) were calculated for every genotype. The resulting data were analyzed as obtained from a randomized complete block design. Significant differences among genotypes were observed for all drought tolerance indices except for SSI and YSI. High yield value in non-stress and stress environments was exhibited by genotypes ‘LR4 × LR25’ (14.02 g) and ‘LR25 × C100’ (3.84 g) respectively. The maximum value of STI (0.93), MP (8.72), GMP (6.91) and HM (5.48) indices was by genotype ‘LR4 × LR25’. Correlation coefficients revealed that TOL, MP, GMP, STI, HM, and YI indices could effectively be used for screening of drought tolerant genotypes. Using MP, GMP, HM, TOL, YI and STI indices, genotypes UPGMA classification was done and three clusters were established that paralleled the biplot analysis results. According to results in this study, ‘LR4 × LR25’ is the most drought tolerant genotype which was clustered as group A. We suggest that tolerance indices including MP, GMP and HM are suitable for sunflower drought tolerant genotypes selection. Additional key words: biplot; diallel cross; Helianthus annuus L.; multivariate analysis; water-stressed condition, yield stability.


Resumen
Evaluación de la reacción de líneas puras de girasol y sus híbridos F 1 a las condiciones de sequía mediante diversos índices de tolerancia al estrés En este estudio, 21 genotipos de girasol (Helianthus annuus L.) derivados de un cruce medio dialélico entre seis líneas puras se evaluaron con y sin estrés hídrico. Para cada ambiente, los genotipos se evaluaron usando un diseño de bloques completos al azar con tres repeticiones. De los datos de rendimiento de grano se calcularon los índices de tolerancia a la sequía para cada genotipo: índice de estabilidad de la tolerancia sexual (ITS), productividad media (MP), productividad media geométrica (GMP), media armónica (HM), índice de susceptibilidad al estrés (SSI), índice de tolerancia (TOL), índice de rendimiento (YI) e índice de estabilidad del rendimiento (YSI). Los datos resultantes se analizaron como obtenidos a partir de un diseño de bloques completos al azar. Se observaron diferencias significativas entre los genotipos en todos los índices de tolerancia a la sequía a excepción de SSI y YSI. Los genotipos 'LR4 × LR25' y 'LR25 × C100' exhibieron un alto rendimiento en ambos ambientes (14,02 y 3,84 g, respectivamente); 'LR4 × LR25' presentó el valor máximo de los índices STI (0,93), MP (8,72), GMP (6,91) y HM (5,48). Los coeficientes de correlación revelaron que los índices TOL, MP, GMP, STI, HM y YI podrían ser utilizados para la selección de genotipos tolerantes a la sequía. Utilizando los índices MP, GMP, HM, TOL, YI y STI, se hizo una Introduction Sunflower (Helianthus annuus L.) is one of the most important oil crops due to its high content of unsaturated fatty acids and a lack of cholesterol (Razi and Assad, 1998). Although sunflower is moderately tolerant to water stress, its production is greatly affected by drought conditions (Pasda and Diepenbrock, 1990). Evidence indicated that drought stress during vegetative phase, flowering and seed filling period causes considerable decrease in yield and oil content of sunflower (Razi and Assad, 1999;Ali et al., 2009). Several research f indings show that plant height, leaf area index and number of green leaves were reduced with no irrigation (Vivek and Chakor, 1992;Agele, 2003;Turhan and Baser, 2004;Poormohammad Kiani et al., 2009). D'Andria et al. (1995 concluded that sunflower yield components such as seed weight is high in irrigation environment compared to non-irrigated conditions. Vasiliu (1988) reported a high seed yield ranging from 2.64 t ha -1 in non irrigated condition to 3.34 t ha -1 with irrigation at 50% field capacity. Similarly, Vijay (2004) mentioned that achene yield in irrigated conditions is higher than non-irrigated one. One of the sunflower breeding goals is establishing resistance to environmental stresses such as drought stress. For this reasons improved tolerance to drought has been a goal but unfortunately, success in breeding for drought tolerance has been limited because (I) it is controlled by several genes, and their simultaneous selection is difficult (Richards, 1996;Yeo, 1998;Flowers et al., 2000); (II) tremendous effort is required to eliminate undesirable genes tightly linked to favorable ones, that are also incorporated during breeding (Richards, 1996); and (III) there is a lack of efficient selection procedures particularly under field conditions (Ribaut et al., 1997;Kirigwi et al., 2004). Generally, different strategies have been proposed for the selection of relative drought-tolerant and resistant genotypes, as Fisher and Maurer (1978) reported that achene yield in drought environment could be considered as drought-resistance index. While Blum (1988) mentioned that selection of genotypes for drought resistance must be associated with selection for high yield in non-stress environments. Hence, by calculation of genotypes yield in drought and well-watered environments, one could select resistant genotypes to drought. There are several selection indices for screening drought resistance genotypes such as geometric mean productivity (GMP) (Fernández, 1992), mean productivity (MP) (Rosielle and Hamblin, 1981), harmonic mean (HM) (Jafari et al., 2009), stress susceptibility index (SSI) (Fischer and Maurer, 1978), yield stability index (YSI) (Bouslama and Schapaugh, 1984), yield index (YI) (Gavuzzi et al., 1997), stress tolerance index (STI) (Fernández, 1992) and tolerance index (TOL) (Rosielle and Hamblin, 1981) that identify susceptible and resistance genotypes based on their yields in stress and non-stress environments. The best selection index must be able to distinguish genotypes that have uniform superiority in both stress and nonstress environment. Fernández (1992) reported that mungbean (Vigna radiata L.) genotypes selection based on STI and GMP indices resulted in genotypes that have high tolerance and high yield. Clarke et al. (1992) used SSI index to distinguish between wheat (Triticum aestivum L.) genotypes. According to Sio-Se Mardeh et al. (2006), MP, GMP and STI were best indices under moderate stress in wheat. The objectives of present study were evaluation of several drought tolerance indices as well as to identify drought-tolerant genotypes in sunflower. clasificación UPGMA de los genotipos y se establecieron tres grupos análogos a los resultados del análisis biplot. De acuerdo a los resultados de este estudio, 'LR4 × LR25', del grupo A, es el genotipo más tolerante a la sequía. Se concluye que los índices de tolerancia MP, GMP y HM son adecuados para la selección de genotipos tolerantes a la sequía en girasol. Palabras clave adicionales: análisis multivariado, biplot; cruce dialélico; estabilidad del rendimiento; estrés hídrico; Helianthus annuus L. basis of their contrasting responses to water stress and different agronomical characteristics revealed in our previous experiments (Poormohammad Kiani et al., 2007a,b, 2008). The RILs used in this experiment were F 9 pure lines which were developed through single seed descent from F 2 plants derived from a cross between 'PAC2' and 'RHA266'. RHA266 was obtained from a cross between wild H. annuus and peredovik by USDA and PAC2 is an INRA-France inbred line from a cross between H. petiolaris and 'HA61' (Gentzbittel et al., 1995). This public RILs population is widely used for genetic analysis of complex traits in sunflower (Poormohammad Kiani et al., 2007a,b, 2008Darvishzadeh et al., 2007;Abou Al Fadil et al., 2007). The six genotypes were grown and crossed in a diallel mating system without reciprocals to produce 15 F 1 hybrid combinations. The parental genotypes and their F 1 hybrids were grown in the greenhouse under controlled conditions. Plants were individually grown in plastic pots containing a mixture of 40% soil, 40% compost and 20% sand as described (Poormohammad Kiani et al., 2007a,b). Temperature was maintained at 25/18 ± 2°C (day/night) and relative humidity at about 65/85 ± 5%. Supplementary light was provided to obtain 16-h light period. Twenty-one genotypes including 15 F 1 hybrids plus 6 parental lines were evaluated in well-watered and water-stressed conditions. In each condition, the genotypes were evaluated using a randomized complete block design with three replications. In order to simulate natural water deficit conditions similar to field, a progressive water stress from mild stress to severe stress was imposed on 45-day-old plants at stage near flower bud formation (R 1 ) (Schneiter and Miller, 1981) for a period of 12 days (Poormohammad Kiani et al., 2007a,b). Both well-watered and water-stressed plants were weighed and water lost replaced carefully. Well-watered (control) plants received sufficient water to maintain soil water content close to field capacity. Water-stressed plants were subjected to a progressive water stress and irrigated with a water volume of 60%, 50% and 40% of field capacity (each 4 days) during 12 days and continued up to harvest. Plants were harvested at maturity, and then the grain yield was recorded for every plot. The drought tolerance indices were calculated for every genotype using the corresponding well-watered and water-stressed plots. The resulting data were analyzed as obtained from a randomized complete block design. Drought tolerance indices were calculated using the equations cited in Table 1.

Statistical analysis
The data were analyzed using the general linear model (GLM) procedure in the SAS software (SAS Institute Inc., Cary, NC, USA). Correlations between grain yield per plant in each of the water regimes and drought tolerance indices were determined using SAS PROC CORR. The Student-Newman-Keuls test was used for comparing mean performance of parents and F1s. Multivariate statistical analysis as principle component analysis, biplot display, three dimensional plots, and cluster analysis were performed using the SPSS version 15.0 and STAT GRAPH softwares.

Results
There were significant differences among genotypes for yield under stress and non-stress conditions (Table 2). Significant differences among genotypes were observed for all drought tolerance indices except for SSI and YSI (Table 2). These results indicate that there is high genetic variation among genotypes, which could be an useful resource for selection of drought-tolerant germplasm. The experimental coefficient of variation (CV) varied from 18.45 to 30.92. However, for the majority of traits the values were less than 25% (Table 2). High yield value in non-stress and stress environments was exhibited by genotypes 'LR4 × LR25' (14.02 g) and 'LR25 × C100' (3.84 g) respectively (Table 3). The maximum value of STI (0.93), MP (8.72), GMP (6.91) and HM (5.48) indices was by genotype 'LR4 × LR25'. The highest value for YI (1.27) was from genotypes 'LR25 × C100' and 'LR4 × LR25' (Table 3). Genotype 'LR4 × LR25' had desirable yield in both environments and hence it is recommendable in arid and semiarid regions. A suitable index must have a significant correlation with yield in any of the two environments. To identify the best index of selection for drought-resistant genotypes, correlation coefficient between these indices and yield in non-stress condition (Y P ) as well as yield in stress condition (Y S ) was determined (Table 4). Correlation coefficients matrix (Table 4) revealed that TOL, MP, GMP, STI, HM, and YI indices could effectively be used for screening of drought resistant genotypes.
Principal component analysis (PCA) revealed that the first PCA explained 72.1% of the total data variation and had positive correlation with the performance under both stress and non-stress environments (Table 5).
Thus the first dimension represents the yield potential and drought tolerance. In other words, this component was able to separate the genotypes with higher yield under both stress and non-stress conditions. The second PCA explained 26.35% of total data variation that have positive and high correlation with SSI but negative one by yield under stress conditions (Table5). Therefore it was the stress susceptibility dimension and was able to separate the susceptible genotypes. Hence, selection of genotypes that have high PCA1 and low PCA2 are suitable for both stress and non-stress condition.
Biplot presentation depicted genotypes 'RHA266 ×  (1981) Stress tolerance index Fernández (1992) Yield index Gavuzzi et al. (1997) Yield stability index Bouslama and Schapaugh (1984) 1 Y S and Y P are stress and optimal (potential) yield of a given genotype, respectively. Y¯S and Y¯P are average yield of all genotypes under stress and optimal conditions, respectively. cated adjacent to important drought resistance indices that confirm these genotypes being drought resistant (Fig. 1). Genotype 'LR4 × C100' was near to SSI and has high Y P (seed yield in non-stress condition) value ( Fig. 1). Therefore, this genotype had specific adaptability to non-stress environment. Some genotypes such as 'LR55 × C100', 'RHA266', 'C104 × C100', 'C104', 'LR4', 'RHA266 × LR25', 'LR55' and 'C100' belong to low yield and high drought sensitivity region in the biplot space (Fig. 1). On the other hand, there was genetic variability among genotypes based on their drought resistance. Using important resistance indices comprising of MP, GMP, HM, TOL, YI and STI genotypes UPGMA classification was done and three clusters were established that paralleled the biplot analysis results (Fig. 2). To identify the relationship among Y P , Y S and suitable resistance indices, three-dimensional graphs for Drought tolerance assessment in sunflower using various stress tolerance indices 1041  each one were also employed (Fig. 3). These graphs showed the ability of these indices to detect Fernández (1992) groups. By using these indices and Y P and Y S variables, three dimensional diagrams could partition the genotypes in four groups: (1) Genotypes producing high yield under both water stress and non-stress environments (group A), (2) genotypes with high yield under either non-stress (group B) or (3) stress (group C) environments, and (4) Table 2.  Figure 1. The genotype by trait biplots of sunflower breeding for resistance to drought stress trial. The traits are spelled out in lowercase letters, and each genotype is represented by numbers. Y P : yield of a given genotype in optimal (potential) conditions. Y S : yield of a given genotype in stress conditions. SSI: stress susceptibility index. TOL: tolerance index. MP: mean productivity. GMP: geometric mean productivity. STI: stress tolerance index. HM: harmonic mean. YI: yield index. YSI: yield stability index.

Discussion
The CV values for yield in non-stress condition (Y P ) and yield in stress condition (Y S ) were 20.96 and 19.92, respectively. Concerning to calculated indices the values varied from 18.45 to 30.92 (Table 2). In general, CV value higher than 20% is considered to be high; however, may be possible to ignore from approximately high CV values when F test are significant and this item is found in several published research works (Takemoto et al., 1988;Xu et al., 2000;Aliyu and Awopetu, 2005;Zarei et al., 2007;Okwuagwu et al., 2008;Kandiç et al., 2009;Sabu et al., 2009). This however indicates that effect of genotypes was more pronounced on studied characters under two irrigation regimes (Aliyu and Awopetu, 2005). The inconsistent CV values reported in many studies as our one might be due to physio-genetic characteristics and degree of compatibility of the plant material, low number of individual per genotype in plot, low number of replication per genotype and/or variable environments in which the trial was carried out (Okwuagwu et al., 2008).
As shown in Table 3, selection based on SSI index identified genotypes with relatively high Y P but low Y S and this is in agreement with Sio-Se Marde et al. (2006) findings in wheat. For the SSI index, the greater values correspond to greater susceptibility of genotype to stress. The main disadvantage of this index is that separated genotypes are not co-located in group A. Clarke et al. (1992) reported that SSI index does not differentiate between potentially drought resistant genotypes and similar limitation was also observed by White and Singh (1991) and Farshadfar and Sutka (2002). YI index that was proposed by Gavuzzi et al. (1997), was significantly correlated with stress and non-stress yield and therefore is a suitable criterion for drought resistance selection. However, Sio-Se Marde et al. (2006) reported that this index groups genotypes only based on their yield under stress condition and so does not discriminate genotypes belonging to group A. In this experiment, a reason for this significant correlation between yield in both conditions and this index may be due to less severe stress conditions. According to Tezara et al. (2002), 67% of field capacity is considered as mild stress and 33% is considered as severe stress in sunflower. In this study, YSI index was not a desirable index because it had no significant relationship with both growing condition's yield. Stability is more important than high yield in non-stress conditions. This index shows the relative stability of yield with conditions changes and the higher the YSI, the more resistant the crop is. Selection based on TOL was eff icient in improving yield under stress condition whereas the selected genotypes performed poorly under non-stress condition (Fernández, 1992). MP is mean yield for a genotype in two stress and non-stress conditions. MP can identify genotypes with high Y P but with relatively low Y S (group B) and thus it fails to distinguish group A from group B. By decreasing TOL and increasing MP, the relative tolerance increases (Rosielle and Hamblin, 1981;Fernández, 1992). GMP is more powerful than MP in separating group A genotypes and has a lower susceptibility to different amounts of Y S and Y P . Therefore MP, which is based on arithmetic mean, will be biased when the difference between Y S and Y P is high. The geometric mean is often used by breeders interested in relative performance since drought stress can vary in severity in field environments and over years (Fernández, 1992). A high STI demonstrates a high tolerance and the best advantage of STI is its ability to separate group A from others. STI index is calculated based on GMP index and therefore there is high positive correlation between these indices (0.963) and agree with Fernández (1992) and Mozaffari et al. (1996). The observed relationships between Y P , Y S and a few desirable criteria (GMP, MP and STI) are consistent with those reported by Fernández (1992) in mungbean (Phaseolus aureus Roxb.), Farshadfar and Sutka (2002) in maize (Zea mays L. ) and Kristin et al. (1997) in common bean (Phaseolus vulgaris L.).
Results based on PCA and biplot analysis revealed that genotype 'LR4 × LR25' classified as group A is a drought resistant genotype that has reasonable yield in both stress and non-stress conditions. PCA and biplot application for screening resistant genotypes agreed with several studies by Fernández (1992), Farshadfar and Sutka (2003) and Golabadi et al. (2006). Result of UPGMA clustering method, based on desirable resistance indices, was in agreement with the output of biplot analysis and therefore by using genotypes that are located in separate groups and have maximum genetic distance, it is possible to analyze genetic parameters of these drought resistance indices.
Drought is considered as one of the most visible factors which affect grain yield and some of the consti- tuents of the seed oil of sunflower (Razi and Assad, 1999;Ali et al., 2009). According to results in this study, the following suggestions are made: 1. In order to select a genotype with stable and high yield in nonirrigated and irrigated conditions, STI, MP, GMP, and HM are proposed as the more suitable indices. Selection by these indices can be useful to identify a cultivar with desirable yield in both stress and non-stress conditions (group A), although it is recommended that, the selection be done based on PCA results (namely by using several indices instead of only one index information). 2. In order to identify parents with high level of differences for confirmation of mapping population for QTL analysis, we suggest making a first selection according to MP, GMP, HM indices to have a maximum of segregating loci.