doi:10.14720/aas.2021.117.1.1656 Original research article / izvirni znanstveni članek Discrimination of drought tolerance in a worldwide collection of safflow-er (Carthamus tinctorius L.) genotypes based on selection indices Pooran GOLKAR 1 2 3, Esmaeil HAMZEH 4, Seyed Ali Mohammad MIRMOHAMMADY MAIBODY 4 Received May 11, 2020; accepted February 28, 2021. Delo je prispelo 11. maja 2020, sprejeto 28. februarja 2021. Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices Abstract: Improvement of elite safflower genotypes for drought-tolerance is hampered by a deficiency of effective selection criteria. The present study evaluated 100 genotypes of safflower in terms of their drought tolerance over a period of three years (2016-2018) under both non-stress and drought-stress conditions. The eight drought-tolerance indices of tolerance index (TOL), mean productivity (MP), geometric mean productivity (GMP), stress susceptibility index (SSI), stress tolerance index (STI), yield stability index (YSI), drought resistance index (DI), and harmonic mean (HARM) were calculated based on seed yield under drought (Y) and non-drought (Yp) conditions. A high genetic variation was found in drought tolerance among the genotypes studied. The MP, GMP, and STI indices were able to discriminate between tolerant and drought-sensitive genotypes. Plots of the first and second principal components identified drought-tolerant genotypes averaged over the three study years. Cluster analysis divided the genotypes into three distinct groups using the drought tolerance indices. Ultimately, eight genotypes (namely, G3, G, G, G24, G33, G47, G58, and G ) from different origins were detected as more tolerant to drought stress suitable for use in safflower breeding programs in drought-affected areas. The most tolerant and susceptible genotypes could be exploited to produce mapping populations for drought tolerance breeding programs in safflower. Key words: cluster analysis; drought stress; principal component analysis; selection index; yield; safflower Abbreviations: TOL: Tolerance; MP: mean productivity; SSI: drought susceptibility index GMP: geometric mean productivity; YSI: yield stability index; DI: drought resistance index. Odkrivanje tolerance na sušo v mednarodni zbirki genotipov žafranike (Carthamus tinctorius L.) na osnovi izbranih indeksov Izvleček: Izboljšanje elitnih genotipov žafranike na prenašanje suše ovira pomanjkanje učinkovitih selekcijskih kriterijev. V raziskavi je bilo ovrednoteno 100 genotipov žafranike glede na njihovo prenašanje suše v obdobju treh let (2016-2018) v razmerah brez stresa in razmerah sušnega stresa. Izračunanih je bilo osem indeksov tolerance na sušni stres kot so tolernca na stres (TOL), poprečna produktivnost (MP), geometrijska poprečna produktivnost (GMP), indeks stresne občutljivosti (SSI), indeks stresne tolerance (STI), indeks stabilnosti pridelka (YSI), indeks odpornosti na sušo (DI), in harmonično poprečje na osnovi pridelka semena v sušnih (Ys) in nesušnih (Yp) razmerah. Med preučevanimi genotipi je bila ugotovljena velika genetska variabilnost v toleranci na sušo. Z indeksi MP, GMP, in STI je bilo mogoče razlikovati na sušo tolerantne in občutljive genotipe. Polja prve in druge glavne komponente so določila na sušo tolerantne genotipe v vseh treh letih raziskave. Klasterska analiza je z uporabo indeksov tolerance na sušo razdelila genotipe v tri jasno ločene skupine. Na koncu je bilo ugotovljenih osem genotipov (G^ ^ G^ ^ in G6i) različ- nega izvora, ki so bili bolj tolerantni na sušo in so primerni za uporabo v žlahtniteljskih programih žafranike na od suše ogroženih območjih. Na sušo najbolj prilagojene genotipe žafranike bi lahko uporabili za odkrivanje populacij, ki bi bile primerne pri žlatnjenju žafranike na sušo. Ključne besede: klasterska analiza; sušni stres; analiza glavnih komponent; selekcijski indeks; pridelek; žafranika Okrajšave: TOL: toleranca; MP: poprečna produktivnost; SSI: indeks občutljivosti na sušo; GMP: geometrijska poprečna produktivnost; YSI: indeks stabilnosti pridelka; DI: indeks odpornosti na sušo. 1 Isfahan University of Technology, Department of Natural Resources, Isfahan, Iran 2 Isfahan University of Technology, Research Institute for Biotechnology and Bioengineering, Isfahan, Iran 3 Corresponding author, e-mail: poorangolkar@gmail.com; golkar@cc.iut.ac.ir 4 Isfahan University of Technology, College of Agriculture, Department of Agronomy and Plant Breeding, Isfahan, Iran Acta agriculturae Slovenica, 117/1, 1-11, Ljubljana 2021 P. GOLKAR et al. 1 INTRODUCTION Droughts due to alterations in rainfall patterns and climate change form a most devastating factor in food production on a global scale (Blum, 2018; Anjum et al., 2017). This warrants a Blue Revolution in agriculture concentrated on increasing productivity per unit of water to produce more crops per drop of water. Recently, an important target in crop breeding programs is the development of drought-tolerant genotypes that possess a high capability for adaption to arid and semi- arid climates (Kirigwi et al., 2004; Basu et al., 2016). The challenges in understanding the mechanisms involved in plant behavior under water scarcity include: 1) mutagenic control of drought tolerance, 2) genetic variability and differences among species in responding to changes in water availability, and 3) interactions with other factors such as drought stress duration and intensity (Varshney et al., 2018). Moreover, breeding programs are adversely affected by the high interaction of genotype x environment, low heritability of drought tolerance traits, lack of efficient selection particularly under field conditions, and the difficulties associated with simultaneous selection, sharp climate changes, and unpredictable rainfall in different regions (Ashraf, 2010; Rauf et al., 2016; Blum, 2018). Since the genotypes with a high yield under optimum conditions may not be drought tolerant (Blum, 2018), many studies preferred selection under both stress and non-stress conditions (Fernandez, 1992). To have a high and durable yield in a drought-prone environment drought-tolerant genotypes are needed (Abdolshahi et al., 2012). The capacity of genotypes to perform reasonably well in drought-stressed environments is the paramount reason for their stable production (Raman et al., 2012). To decrease the impacts of abiotic stress without any substantial yield loss, researchers tend to develop drought-tolerant genotypes based on prior evaluation and identification of drought-tolerant germplasm. The high cost of drought soil amelioration has encouraged breeders to use selection indices as an economic and efficient method for resolving the problems associated with drought stress breeding (Vieira et al., 2016). In this regard, a variety of selection indices to identify stress-tolerant cultivars have been proposed that some of the important and most applicable of them include: Tolerance (TOL) (Rosielle $ Hamblin, 1981) (Table 1) , mean productivity (MP) (Rosielle & Hamblin, 1981) (See Table 1), stress susceptibility index (SSI) (Fischer and Maurer, 1978) (See Table 1), geometric mean productivity (GMP) (Kristin et al., 1997) (See Table 1), stress tolerance index (STI) (Fernandez, 1992) (See Table 1), yield stability index (YSl) (Gavuzzi et al., 1997) (See Table 1), and drought resistance index (DI) (Lan, 1998) (See Table 1). Our literature review have reported on the efficiency of different selection indices for selecting drought-tolerant genotypes in different crops as like as rice (Raman et al., 2012); canola (Khalili et al., 2012);sunflower (Gholinezhad et al., 2014); maize (Hao et al., 2011) and bread wheat (Abdolshahi et al., 2012). Safflower (Carthamus tinctorius L.) is an annual oil Table 1: Different drought tolerance indices used for screening safflower genotypes Index name Equation- Relereuce Mean productivity ys + rp MP = p RoskUe and Hamblin, i'jsi Tolerante index (TOL) II x TS •K tA Ficsher and Maurer, 1978 Geometrie Mean Productivity (GMP) CMP - Yp}(Ys) Krislin el al., 1997 Stress Susceptibility Index (SSI) "00 Rosielle and Hamblin, 1981 Stress Tolerance Index (STI) YsxYp STI = ——-Yp1 Fernandez, 1992 Yield Stability Index (YSl) Ys YSl ~ — Y„ Gavuzzi etal,, 1997 Drought Resistance Index (DI) .-"SS Ys Lan, 1998 Harmonie Mean HARM = 2(YfX Ys)/(Yp + Ys) Krislin el al., 1997 10 Acta agriculturae Slovenica, 117/1 - 2021 Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices seed crop with diverse industrial and pharmaceutical application that is grown commercially in Iran (Golkar & Ka-rimi, 2019). The deep roots of safflower make it a drought-tolerant plant viable under the drought stress conditions in arid climates (Mirzahashemi et al., 2014; Hussain et al., 2016). Drought stress is one of the most devastating abiotic stresses that poses a serious threat to worldwide safflower production (Hussain et al., 2016). Given the declining water resources in the arid and semi- arid regions of the world due to consecutive droughts, increased safflower cultivation can be an economic and valuable alternative to other drought-tolerant genotypes. In this regards, some studies is known about drought tolerance of local Iranian cultivars (Omidi et al., 2012; Bahrami et al., 2014; Mirzahashemi et al., 2014). Despite of current efforts intended for assessing tolerance criteria based on tolerance indices in safflower, little has been reported at maturity (Bahrami et al., 2014). Furthermore, this tolerance undoubtedly appears to be stage-dependent and must be evaluated at the yielding phase. Variations in drought patterns such as differences in location, year, and drought intensity as well as genotypic differences call for safflower genotypes with different levels of drought tolerance to be cultivated in different areas. However, the differences in the genotypes recommended might have stemmed from the variability in the drought tolerance potential of safflower genotypes. Moreover, climate changes increase drought frequency in some regions but drought is a global issue. Given the broad distribution of safflower around the world, it is the objective of the present study to identify drought-tolerant genotypes from a new worldwide collection based on drought selection indices. The new identified genotypes could be used for cultivation in arid regions of the world. 2 MATERIALS AND METHODS 2.1 PLANT MATERIAL One hundred safflower genotypes originating from different geographical regions of the world were selected for screening drought tolerance (Table S1). The exotic genotypes were obtained from Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Germany. Iranian genotypes were taken from the genotype inventory at the Agricultural Research Center, Isfahan, Iran. 2.2 FIELD EXPERIMENT AND IRRIGATION REGIMES An experiment was carried out in three consecutive years from early March 2016 to the end of 2018 at Lavark Research Farm, affiliated to Isfahan University of Technology, 40 Km southwest of Isfahan (32° 32'N, 51° 23' E, 1630 m above sea level), Iran. Mean annual precipitation and temperature at this site are 149 mm and 15.4°C, respectively. The soil was silty clay loam with a bulk density of 1.3 g cm3 in the top 50 cm and a pH level of 7.4-7.9. The field experimental design was a square lattice design (10 by 10) with two replications for each different irrigation (drought stress and non-drought) regimes in each year. The seeds were planted in rows of 3 m long and spaced 25 cm from each other to yield a plant density of 40 plants m2 in the plots. All the plants received the first irrigation before planting. After this period, irrigation was applied every week until the budding stage. From budding stage to full maturity stage, the non-stress treatment involved irrigation when 50 % of the total available water was depleted from the root zone, but in the drought stress conditions, irrigation was applied when 85 % of the total available water was depleted from the root zone. The irrigation interval (number of days between two irrigations) during the growing season (budding to full maturity) was variable because of the variation in evapotranspiration (ET). Soil samples were taken from a depth of 0 to 60 cm of the soil from both drought and non- drought plots to determine the soil water content and calculate the irrigation water content on the basis of a 60 cm rooting depth. Soil samples were taken before each irrigation when evaporation from a Class A pan indicated 70 and 140 mm of evaporation under normal and drought-stress conditions, respectively. Then, irrigation depth was determined using the formulae: I = [(0 - 0.)/100] D x B); where, I represents irrigation depth (cm), 0FC (-0.03 MPa) is soil gravimetric moisture percentage at field capacity (22 %), 0i (-1.5 MPa) is soil gravimetric moisture percentage at irrigation time (10 %), D is root-zone depth (50 cm), and B is soil bulk density at the root zone (1.3 g cm-3) (Clarke et al., 2008). The volume of irrigation water applied was monitored at each irrigation by calculating the depth of water over a Parshall flume which was calculated as: Id = I xp, where p is the fraction of I that can be depleted from the root zone. Then Ig = (Id/ Ea) x 100, which Ea is irrigation efficiency (%), assumed to be 65 % on the average. The differences in available water related to different mean for temperature in growing seasons across three years of study. No growth regulators or fungicides were applied. Surface application of 130 (kg ha1) N and 25 (kg ha1) P was carried out in both treatments with an additional 55 kg ha1 of N during the rosette stage. Plants were harvested in the middle row at maturity and seed yield was recorded in each plot. Ten different selection indices were calculated using the equations reported in Table 1. In these equations, Y"S represents the Acta agriculturae Slovenica, 117/1 - 2021 11 P. GOLKAR et al. yield of genotypes under stress; Yp the yield of genotypes under normal conditions (kg ha"1); and denote mean yields of all the genotypes under stress and non-stress conditions, respectively. 2.3 STATISTICAL ANALYSIS A combined analysis of variance (ANOVA) was performed using SAS software (SAS. Ver. 9.3.1), for seed yield and selection indices using GLM procedure. Principal Component Analysis (PCA) and 3D biplot diagrams were exploited to identify tolerant and susceptible genotypes using R software (Ver. 3.6.1). Correlations between seed yield in the non-stress and drought-stress treatment as well as the relevant drought tolerance indices were determined using SAS PROC CORR and Heat Map Graph (R software ver 3.6.1). The safflower genotypes were classified using the seed yields obtained from each of the water treatments and drought tolerance indices data using the Ward algorithm based on the squared Euclidean distances in the R Software (Ver. 3.6.1). 3 RESULTS Analysis of variance indicated the non-significant effect of year on all the studied traits (Table 2). A highly significant (p < 0.01) variation in seed yield was observed among the studied genotypes under both (stress and non-stress) conditions and for all the tolerance indices examined (Table 2). The genotype x year interaction effect was not significant for any of the indices, except for DI and HARM (Table 2). The significant genotype x environment interaction for both DI and HARM, indicating considerable variability among the genotypes across different years and different irrigation treatments for these selection indices. Table 3 reports the ten highest and the lowest seed yields, Yp and Ys, for the studied genotypes. Clearly, the highest Yp values were obtained for G13 (5680 kg ha-1) (from Iran) and G61 (5310 kg ha1) (from Morocco), but the least Yp value was obtained for G79 (900 kg ha"1). Under stress conditions, the highest seed yield (Ys) was obtained in genotypes 47 (3038 kg ha"1) and 24 (2670 kg ha"1), but the lowest (590 kg ha"1) was observed in G79 (Table 3). G13 recorded the highest values of TOL (4130), SSI (1.61), and HARM (1.14) indices, whereas G86 (from Tajikistan) recorded the least values for TOL (2608)6, SSI (0.24), and HARM (0.13). The highest (0.87) and the lowest (0.27) values of YSI were obtained for G86 and G13, respectively. Finally, the genotypes 47 and 25 had the highest (1.55) and lowest (0.20), respectively, mean values of the DI index (Table 3). The correlation coefficients among Yp, Ys, and other drought tolerance selection indices were calculated to determine the most desirable drought tolerance criteria (Table 4). It was found that seed yield and YSI exhibited negative (- 0.5**) and positive (0.34**) correlations under the non-stress and stress conditions, respectively. Seed yield under the non-stress treatment showed positive and significant correlations with all the selection indices, except for YSI ("0.50**) and DI (Table 4). Seed yield under the stress treatment showed positive and significant correlations with MP, GMP, STI, and DI but negative and significant ones with SSI and HARM (Table 4). Principal Component Analysis (PCA) as a representative for distinguish the relationships among the indices revealed that the first component (PC1) explained 54 % of the total seed yield variation and exhibited positive correlations with Yp, MP, GMP, and STI (Figure 1). PC2 explained 44 % of the total yield variation and had higher positive correlations with DI, YSI, and Ys but higher negative correlations with SSI and HARM (Figure 1). Table 2: Combined analysis of variance for seed yield under non-stress (Yp) and stress (Ys) conditions and different susceptibility indices in safflower genotypes growing under drought stress and normal conditions evaluated in 2016 and 2018 S.O.V DF yp Ys SSI YSI TOL MP GMP STI DI HARM Year (Y) 2 3947374.2 77841 0.00019 0.05 133849.6 29145.2 303800.5 0.047 0.34 0.2 Block/ Year 3 282227.2 117545.25 0.36 0.06 696640.8 25726.0 14486.5 0.003 0.13 0.18 Genotype 99 (G) 3305183.4** 844416.53** 0.40** 0.08** 2177268.3** 1530482.9** 1260048.7** 0.53** 0.34 0.23** G x Y 198 89064.9 80535.49 0.071 0.01 143568.0 48908.2 55476.1 0.028 0.05** 0.54** Residual 297 89064.9 154372.9 0.12 0.025 350774.0 122390.6 125950.5 0.06 0.09 0.07 * and **, significant at p < 0.05 and p < 0.01, respectively. Abbreviations: DF: degree of freedom; Yp: seed yield under non-stress; Ys: seed yield under stress; SSI: stress susceptibility index, YSI: yield stability index; TOL: stress tolerance; MP: mean productivity; GMP: geometric mean productivity; STI: stress tolerance index; DI: Drought Resistance Index; and HARM: Harmonic mean 10 Acta agriculturae Slovenica, 117/1 - 2021 Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices Table 3: Ten highest and lowest values for seed yield under non-stress conditions (Yp), Seed yield under stress conditions (Ys), and different selection indices among the 100 different safflower genotypes investigated Ten highestYP¥ indices (kg ha-1) YS (kg ha-1) SSI YSI TOL MP GMP STI DI HARM 5680(G13) 3080(G47) 1.613(G13) 0.8724(G86) 4130(G13) 3985(G47) 3820.2(G47) 2.2172(G47) 1.5516(G47) 1.1488(G13) 5310(G61) 2670(G24) 1.58(G61) 0.8552(G50) 3770(G61) 3615(G13) 3233.2(G24) 1.6356(G11) 1.3097(G24) 1.1104(G61) 4910(G58) 2390(G11) 1.5697(G25) 0.84(G99) 3140(G58) 3425(G61) 3159.8(G11) 1.5709(G24) 1.1934(G94) 1.1069(G76) 4890(G47) 2385(G3) 1.5629(G76) 0.8148(G38) 2800(G76) 3380(G11) 3046.4(G3) 1.4443(G3) 1.1672(G97) 1.1029(G25) 4520(G33) 2350(G97) 1.5285(G18) 0.8042(G81) 2760(G33) 3340(G58) 2957.2(G13) 1.33(G13) 1.15(G69) 1.0677(G18) 4370(G11) 2167.5(G48) 1.5057(G59) 0.7948(G94) 2750(G18) 3305(G24) 2931(G58) 1.2916(G58) 1.1144(G48) 1.03(G59) 3980(G27) 2160(G2) 1.4701(G27) 0.7841(G41) 2690(G27) 3142.5(G3) 2846.3(G61) 1.2356(G61) 1.1077(G99) 1.0095(G27) 3980(G18) 2130(G63) 1.4483(G66) 0.7749(G12) 2450(G59) 3140(G33) 2824.3(G97) 1.215(G97) 1.0974(G63) 0.9724(G66) 3940(G24) 2118.3(G94) 1.4284(G58) 0.7713(G69) 2440(G25) 2880(G97) 2817.8(G33) 1.2063(G33) 1.0941(G86) 0.97(G21) 3900(G3) 2100(G69) 1.4258(G21) 0.7668(G37) 2300(G66) 2845(G2) 2759(G2) 1.1386(G2) 1.0793(G11) 0.9508(G58) Ten lowest Yp indices (kg ha-1) ys (kg ha-1) SSI YSI TOL MP GMP STI DI HARM 1 2660(G8) 930(G78) 1.3358(G95) 0.7411(G10) 1915(G39) 2660(G57) 2568.3(G48) 1.0026(G48) 1.0463(G96) 0.8926(G95) 2 1510(G81) 920(G95) 0.5084(G69) 0.3546(G21) 410(G71) 1210(G67) 1167.2(G64) 0.209(G31) 0.3026(G79) 0.2604(G69) 3 1460(G64) 920(G87) 0.4939(G12) 0.3461(G66) 380(G99) 1200(G64) 1161.3(G31) 0.2038(G64) 0.3007(G95) 0.2553(G12) 4 1450(G67) 910(G37) 0.4824(G41) 0.3334(G27) 340(G38) 1175(G71) 1155.7(G71) 0.2032(G71) 0.3003(G66) 0.2536(G41) 5 1380(G89) 890(G21) 0.4545(G94) 0.3204(G59) 339(G5) 1160(G89) 1138.2(G89) 0.1942(G89) 0.2884(G18) 0.2367(G94) 6 1380(G71) 860(G31) 0.4392(G81) 0.3093(G18) 320(G81) 1150(G9) 1071.8(G9) 0.173(G9) 0.2834(G87) 0.2299(G81) 7 1350(G82) 801(G5) 0.4147(G38) 0.2939(G76) 310(G79) 1070(G82) 1046.6(G37) 0.1647(G37) 0.2647(G59) 0.208(G38) 8 1210(G37) 790(G82) 0.3429(G99) 0.2904(G25) 300(G50) 1060(G37) 1026(G82) 0.1582(G82) 0.2497(G21) 0.1803(G99) 9 1140(G5) 780(G9) 0.3284(G50) 0.2884(G61) 300(G37) 970.5(G5) 952.4(G5) 0.1363(G5) 0.2346(G76) 0.1587(G50) 10 900(G79) 590(G79) 0.2846(G86) 0.2719(G13) 260(G86) 745(G79) 718.9(G79) 0.0788(G79) 0.2024(G25) 0.1392(G86) Table 4: Correlation coefficients between seed yield (kg ha ') of safflower genotypes under non-stress (Yp) and stress (Ys) conditions and each of the stress susceptibility indices averaged over three years Y p Y s SSI YSI TOL MP GMP STI HARM DI Y¥ p 1 Y s 0.59** 1 SSI 0.50** -0.34** 1 YSI -0.50** 0.34** -0.99** 1 TOL 0.86** 0.10 0.83** -0.83** 1 MP 0.95** 0.80** 0.24* -0.24* 0.67** 1 GMP 0.88** 0.89** 0.08 -0.08 0.52** 0.98** 1 STI 0.86** 0.88** 0.06 -0.06 0.50** 0.96** 0.98** 1 HARM 0.53** -0.32** 0.99** -0.99** 0.86** 0.27** 0.10 0.08 1 DI 0.17 0.89** -0.70** 0.70** -0.31** 0.47** 0.61** 0.61** -0.67** 1 * and ** Significant at p < 0.05 and p < 0.01; respectively; ns, not significant. Abbreviations ¥: Yp: Seed yield under non- stress condition; Ys: Seed yield under stress condition; SSI: stress susceptibility index, YSI: yield stability index, TOL: stress tolerance, MP: mean productivity, GMP: geometric mean productivity, STI: stress tolerance index, HARM: harmonic mean, DI: drought resistance index. Acta agriculturae Slovenica, 117/1 - 2021 11 P. GOLKAR et al. c + X M Di ¿g(9 - tbi 4 » C.u8 CM G«^ -"O« * 0, W etoeMQW »GÎJOI GiSGO* Q74 Ci * OSJMS •• i * * -r t CM » * « D *an * * CM) M ®M 4 OH6H ZSiGl* -C.O-Qi( Cil C-« * en "tic * « H« ou* *atjv .r , YSI » G7I *«l * B —i a 1 i a p« Figure 1: Biplot drawn based on the first and second components obtained from principal component analysis using seed yield of safflower genotypes under non stress (Yp) and stress (Ys) conditions. Abbreviations: stress susceptibility index (SSI), yield stability index (YSI), stress tolerance (TOL), mean productivity (MP), geometric mean productivity (GMP), stress tolerance index (STi), drought resistance index (Di); harmonic mean (Harm), and conditions in 100 safflower genotypes. 3000 1000 Figure 2: Three-dimensional diagram for identifying drought-tolerant genotypes based on seed yield under non-stress (Yp) and stress (Y ) conditions as well as the stress tolerance index (STi). Because of the positive and significant correlation of STI with seed yield under both conditions, a three-dimensional graphs based on the STi index were drawn (Figure 2). These graphs split the genotypes into four groups, each of which represents one combination of the genotypes. The genotypes 47, 24, 97, 3, and 11 (Group A) are those with high yields under drought and non-stress environments. Those in Group B (e.g., G13, G59, and G61) consisted of genotypes with high yields in a normal environment but low seed yields under drought conditions. No genotype was, however, detected as one with a high yield in a stressful environment (Group C). The genotypes with low yields under both environmental conditions were assigned to Group D (e.g., G3, G, G79, and G5). 3.1 CLUSTER analysis A dendrogram was drawn based on the cluster analysis using seed yield under drought and non-drought conditions along with the selection indices TOL, MP, GMP, STi, SSi, YSi, Di, and Harm (Figure 3). The cluster analysis performed classified the 100 genotypes of safflower investigated into three distinct groups consisting of 7, 44, and 49 genotypes. The genotypes in the smallest group (1) including ^ ^ ^ ^ ^ and G3 showed the highest seed yield under both non-stress and drought stress conditions (Figure 3). The genotypes clustered in Group 2 (i.e., G51, G93, G39, G60, G2, G57, G14, G16, G97, G48, G63, G99, G5Q, G86, G1Q, G65, and G81) rec°rded medium levels of seed yield under drought stress. The 10 Acta agriculturae Slovenica, 117/1 - 2021 Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices Figure 3: Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices. third group consisted of genotypes with a low productivity under either environmental conditions (i.e., G79, G37, Gsg, Grr, G„, and GJ. 4 DISCUSSION This study evaluated drought tolerance in a world collection of safflower accessions under the effects of year and genotype. The analysis of variance showed a large genetic variation in drought tolerance among the accessions studied as an unpredictable factor affecting seed yield in the genotypes from different geographical regions. Year factor was not found to have any significant effect on seed yield or selection indices; hence, the indi- ces selected for this germplasm can be effectively used if seed yield is adequately heritable. Considering the fact that traits involved in drought tolerance mechanisms are polygenic ones, the requirement to screen tolerant genotypes has encouraged plant breeders to seek a reliable index. In response to this need, the present study evaluated eight different selection indices (i.e., MP, GMP, TOL, SSI, STI, YSI, DI, and HARM) for use in the estimation of seed yield under drought stress. Based on the correlation analysis performed, the positive and significant correlation between TOL and Yp (0.86**) implies that the genotypes superior in terms of seed yield (such as G79 and G9) showed greater reductions in seed yield under drought conditions. Also, the non-significant correlation between TOL and Y (0.10) revealed the failure of the TOL index Acta agriculturae Slovenica, 117/1 - 2021 11 P. GOLKAR et al. to identify the most tolerant genotypes, confirming the results reported by Rizza et al. (2004). The greater TOL values indicated the higher sensitivity of the genotypes investigated to drought stress; thus, smaller values of this index is favored. The positive and significant correlation between Y and SSI (0.50**) and that between Y P »» P and HARM (0.53**) demonstrated that the genotypes with higher values for Yp or the SSI index exhibited a higher sensitivity to drought stress (Table 2). On the other hand, the negative and significant correlation between SSI and Ys (-0.34**) or that between HARM and Ys (-0.32**) implied that the superior genotypes under drought stress recorded lower values for SSI and HARM. Hence, the SSI and HARM indices are able to discriminate superior safflower genotypes (the ones with lower values of SSI or HARM indices) in drought prone areas. Studying spring wheat, Guttieri et al. (2001) maintained that SSI values >1 and <1 might indicate above-average and below-average susceptibility to drought stress, respectively. The most suitable index for selecting stresstolerant genotypes is an index that establishes a positive and strong correlation with seed yield under both stress and non-stress conditions (Fernandez, 1992). To select drought-tolerant genotypes, based on the most desirable indices, use is made of the correlation coefficient of each index with Y and Y (Golabadi et al., 2006; Ebrahymian et al., 2012; aabdolshahi et al., 2012; Naghavi et al., 2013). Seed yield was found to have a highly significant positive correlation with GMP, MP, STI, and HARM indices under both the environmental (drought and non- drought conditions) conditions examined (Table 2). Based on the correlation analysis conducted in this study, GMP, STI, and MP were found to favor genotypes with a high-yield potential under stress conditions (Table 2), which agrees with the findings reported Sio-Se Mardeh et al. (2006), Hao et al. (2011), and Ebrahimiyan et al. (2012). Given the fact that G47 recorded the highest values for MP and STI, this genotype was identified as the most productive and stable safflower ones from among the ones investigated under both stress and non-stress conditions. The results of the present study indicating the capability of the selection indices GMP, MP, and STi to identify genotypes satisfactorily under both conditions are consistent with those reported for GMP and MP in mungbean (Fernandez, 1992); STi and GMP in rice (raman et al., 2012); safflower (bahrami et al., 2014) Brassica napus L. (Khalili et al., 2012) and durum wheat (ilker et al., 2011); as well as GMP, STi, and MP in tall fescue (Ebrahymian et al., 2012) and maize (Hao et al., 2011). GMP is often used by plant breeders interested in calculating relative performance since drought stress might vary in severity both under field conditions and over different years (Fernandez, 1992). in the present study, GMP established significant and positive correlations with TOL, Yp, and Ys (Table 3). DI, which is commonly accepted as an index to identify genotypes with high yields under both stress and non-stress conditions (Lan, 1998), showed only a highly significant and positive correlation with Y (0.89**) (Table 4), demonstrating that selection of safflower genotypes with high DI values might be useful for severe drought-stricken regions but that the genotypes selected based on this index do not have very high yields or yields equivalent to those of genotypes currently cultivated under normal irrigation. Seed yield under non-drought conditions (Yp) was positively correlated with Ys, confirming previous reports on safflower (Bahrami et al., 2014) other crop species such as bread wheat (El-Rawy and Hassan, 2014), corn (Naghavi et al., 2013) and bread wheat (Abdolshahi et al., 2012). it may also be noted that the satisfactory responses shown by some genotypes under stress conditions could be ascribed to the good adaptation mechanisms in these genotypes (Naghavi et al., 2013). The impacts of the different indices in each PC indicate that PC1 and PC2 could be identified as yield potential and stress susceptibility groups, respectively. The genotypes (such as G47, G, G3, and G24) recording high values for both PC1 and PC2 may be considered as superior ones for seed yield under both experimental conditions; hence, they are designated as stable genotypes (Figure 1a). The genotypes recording low PC1 but high PC2 values included those also with high values of Di, YSi, and seed yield under drought stress, but low values of SSi and Harm values (Figure 1b). The genotypes (such as G and G) recording high PC1 but low PC2 values included genotypes with high values for GMP, STi, MP, and seed yields under non-stress conditions (Figure 1C). On the other hand, the majority of the genotypes with low PC1 and PC2 values were identified as susceptible genotypes; these included G79, as the most tolerant one, and the genotypes G9, G82, and G89 (Figure 1D), that were recognized as unstable genotypes. The majority of the genotypes investigated (more than 60 %) were classified in Groups B and D (Figure 1). This biplot may also be used for identifying contrasting genotypes (genotypes in group A versus D) for planning fine mapping populations for safflower genome studies of drought tolerance. Based on our cluster analysis, the genotypes assigned to Group 3 were recognized as the most tolerant ones to be used as parents for improving drought tolerance in safflower breeding programs. Thus, the genotypes in Group 1 and Group 3 were identified as drought tolerant and drought susceptible, respectively. Cluster analysis has been widely used not only to discriminate high distance genotypes but also to determine genetic diversity based on similar traits under drought stress conditions (Golabadi et al., 2006; Moham-madi et al., 2011; Naghavi et al., 2013).The results of the 10 Acta agriculturae Slovenica, 117/1 - 2021 Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices present cluster analysis of the genotypes investigated were consistent with the PCA results obtained. Thus, drought-tolerant genotypes recording high PCt and PC2 values as well as those assigned to Groups of 1 and 3 in the cluster analysis can be used as extreme parental genotypes with the highest genetic distance to develop new hybrid varieties in safflower aimed at production of drought-tolerant cultivars. However, further evaluation of genotypes using drought tolerance indices across multiple locations is required to confirm their stability for developing improved safflower genotypes. 5 CONCLUSION From the results obtained, it may be concluded that it is preferable to use simultaneously different drought tolerance indices for screening drought-tolerant safflow-er genotypes. The results of different multivariate analyses revealed that STI, MP, and GMP, in this descending order, were not only capable of efficient selection of high seed-yield genotypes under both the environmental conditions examined but also of discrete identification of drought-tolerant from drought-sensitive safflower genotypes. The G47 genotype (Spanish origin) was identified as the most drought-tolerant one with the highest seed yield under both drought and non-stress conditions. Based on the results obtained in this study, the elite genotypes (i.e., G24, G , G3, G , G33, G58, and G ) may be recommended as promising cultivars for cultivation in drought-affected areas or as appropriate donor parents in safflower hybridization programs. These genotypes may also be exploited for improving seed yield and stability in safflower for cultivation in drought prone regions through appropriate selection methods. 6 ACKNOWLEDGMENT The authors would like to thank Research Institute for Biotechnology and Bioengineering, Isfahan University of Technology, Isfahan, Iran. Research Center, Isfahan, Iran, is also acknowledged for their financial support as a research project number # 97/ 70649. 7 REFERENCES Abdolshahi, R., Safarian, A., Nazari, M., Pourseyedi, S., Mo-hamadi-Nejad, G. (2013). Screening drought-tolerant genotypes in bread wheat (Triticum aestivum L.) using different multivariate methods. Archives of Agronomy and Soil Science, 59(5), 685-704. https://doi.org/10.1080/03650340. 2012.667080 Anjum, S.H.A., Ashraf U., Zohaib A, Tanverr M., Naeem M., Iftikhar A, Tabassum T., Nazir U. (2017). Growth and developmental responses of crop plants under drought stress: a review. Zemdirbyste-Agriculture, 104(3), 267-276. https:// doi.org/10.13080/z-a.2017.104.034 Ashraf, M. (2010). Inducing drought tolerance in plants: Recent advances. Biotechnology Advances, 28,169-183. https://doi. org/10.1016/j.biotechadv.2009.11.005 Bahrami, F., Arzani, A., Karimi, V. (2014). Evaluation of yield-based drought tolerance indices for screening safflower genotypes. Agronomy Journal, 106(4), 1219-1224. https:// doi.org/10.2134/agronj13.0387 Basu, S., Ramegowda, V., Pereira, A. (2016). Plant adaptation to drought stress. F1000 Research, 5, 1554. https://doi. org/10.12688/f1000research.7678.1 Blum, A. (2018). Plant Breeding for Stress Environments. CRC press, 231 pp. https://doi.org/10.1201/9781351075718 Clarke, T.C., Parkin, G.W., Ferre, T.P.A. (2008). Soil water content. In: Carter MR, Gregorich EG, editors. Soil sampling and methods of analysis. Pinawa: Canadian Society of Soil Science. Ebrahimiyan, M., Majidi, M.M., Mirlohi, A., Gheysari, M. (2012). Drought-tolerance indices in a tall fescue population and its polycross progenies. Crop and Pasture Science, 63(4), 360-369. https://doi.org/10.1071/CP11279 FAOSTAT (2016). http://faostat.fao.org. Fernandez, G.C.J. (1992). Effective selection criteria for assessing plant stress tolerance. In: C.G. Kuo, editor, Adaptation of Food Crops to Temperature and Water Stress: Proceedings of an International Symposium, Taiwan. 13-18 Aug. 1992. Asian Vegetable Res. and Dev. Ctr., Shanhua, Tainan, Taiwan. p. 257-270. Fischer, R.A., Maurer, R. (1978). Drought resistance in spring wheat cultivars: 1. Grain yield response. Australian Journal of Agricultural Research, 29, 897-912. https://doi. org/10.1071/AR9780897 Gavuzzi, P., Rizza, F., Palumbo, M., Campaline, R.G., Ricciardi, G.L., Borghi, B. (1997). Evaluation of field and laboratory predictors of drought and heat tolerance in winter cereals. Plant Science, 77, 523-531. https://doi.org/10.4141/P96-130 Gholinezhad, E., Darvishzadeh, R., Bernousi, I. (2014). Evaluation of drought tolerance indices for selection of confectionery sunflower (Helianthus anuus L.) landraces under various environmental conditions. Notulae Botanicae Horti Agrobotanici Cluj-Napoca, 42(1), 187-201. https://doi. org/10.15835/nbha4219394 Golabadi, M., Arzani, A., Mirmohammadi Maibody, S.A.M. (2006). Assessment of drought tolerance in segregating populations in durum wheat. African Journal of Agricultural Research, 5, 162-171. Golkar, P., Karimi S. (2019). Safflower (Carthamus tincto-rius L.) Breeding. Edited by: Al-Khayri, J.M., Jain, S.M. and Johnson, D.V., In: Advances in Plant Breeding Strategies: Industrial and Food Crops. Springer. Pub. https://doi. org/10.1007/978-3-030-23265-8_14 Guttieri, M.J., Stark, J.C., Brien, K., Souza, E. (2001). Relative sensitivity of spring wheat grain yield and quality parameters to moisture deficit. Crop Science, 41, 327-335. https:// doi.org/10.2135/cropsci2001.412327x Acta agriculturae Slovenica, 117/1 - 2021 11 P. GOLKAR et al. Hao, Z.F., Li, X.H., Su, Z.J., Xie, C.X., Li, M.S., Liang, X.L., Weng, J.F., Zhang, D.G, Li, L., Zhang, S.H. (2011). A proposed selection criterion for drought resistance across multiple environments in maize. Breeding Science, 61(2), 101108. https://doi.org/10.1270/jsbbs.61.101 Hussain, M.I., Lyra, D.A., Farooq, M., Nikoloudakis, N., Khalid N. (2016). Salt and drought stresses in safflower: a review. Agronomy and Sustainable Development, 36(1), p.4. https:// doi.org/10.1007/s13593-015-0344-8 Ilker, E., Tatar, O., Aykut Tonk, F., Tosun, M., Turk, J. (2011). Determination of tolerance level of some wheat genotypes to post-anthesis drought. Turkish Journal of Field Crops, 16(1), 59-63. Khalili, M., Naghavi, M.R., Pour Aboughadareh, A.R., Talebza-deh, J. (2012). Evaluating of drought stress tolerance based on selection indices in spring canola cultivars (Brassica napus L.). Journal of Agricultural Science and Technology, 4(11), 78-85. https://doi.org/10.5539/jas.v4n11p78 Kirigwi, F.M., Van Ginkel, M., Trethowan, R., Seaes, R.G., Rajaram, S., Paulsen, G.M. (2004). Evaluation of selection strategies for wheat adaptation across water regimes. Euphytica, 135. https://doi.org/10.1023/ B:EUPH.0000013375.66104.04 Lan, J. (1998). Comparison of evaluating methods for agronomic drought resistance in crops. Acta Agriculturae Bore-ali-occidentalis Sinica, 7, 85-87. Mirzahashemi, M., Golkar, P. and Mohamadinejad, G. (2014). Gene effects for agronomic traits in safflower (Carthamus tinctorius L.) under drought stress. Ethno-Pharmaceutical Products, 1(1), 23-28. Mohammadi, M., Karimizadeh, R. and Abdipour, M. (2011). Evaluation of drought tolerance in bread wheat genotypes under dryland and supplemental irrigation conditions. Australian Journal of Crop Science, 5(4), 487-493. Naghavi, M.R., Aboughadareh, A.P., Khalili, M. (2013). Evaluation of drought tolerance indices for screening some of corn (Zea mays L.) cultivars under environmental conditions. Notulae Scientia Biologicae, 5(3), 388-393. https://doi. org/10.15835/nsb539049 Omidi, A.H., Khazaei, H., Monneveux, P. Stoddard, F. (2012). Effect of cultivar and water regime on yield and yield components in safflower (Carthamus tinctorius L.). Turkish Journal of Field Crops, 17, 10-15. R Core Team. (2008). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Raman, A., Verulkar, S., Mandal, N., Variar, M., Shukla, V., Dwivedi, J., Singh, B., Singh, O., Swain, P., Mall, A., Robin, S. (2012). Drought yield index to select high yielding rice lines under different drought stress severities. Rice, 5(1), 31. https://doi.org/10.1186/1939-8433-5-31 Rauf, S., Al-Khayri, J.M., Zaharieva, M., Monneveux, P. and Khalil, F. (2016). Breeding strategies to enhance drought tolerance in crops. In Advances in plant breeding strategies: agronomic, abiotic and biotic stress traits (pp. 397-445). Springer, Cham. https://doi.org/10.1007/978-3-319-22518-0_11 Rizza, F., Badeckb, F.W., Cattivellia, L., Lidestric, O., Di Fon-zoc, N., Stancaa, A.M. (2004). Use of a water stress index to identify barley genotypes adapted to rainfed and irrigated conditions. Crop Science, 44, 2127-2137. https://doi. org/10.2135/cropsci2004.2127 Rosielle, A.A., Hamblin, J. (1981). Theoretical aspects of selection for yield in stress and non-stress environments. Crop Science, 21, 43-46. https://doi.org/10.2135/cropsci1981.001 1183X002100060033x Sio-Se Mardeh, A., Ahmadi, A., Poustini, K., Mohammadi, V. (2006). Evaluation of drought resistance indices under various environmental conditions. Field Crops Research, 98, 222 - 229. https://doi.org/10.1016/j.fcr.2006.02.001 Varshney, R.K., Tuberosa, R., Tardieu, F. (2018). Progress in understanding drought tolerance: from alleles to cropping systems. Journal of Experimental Botany, 69(13), 3175 -3179. https://doi.org/10.1093/jxb/ery187 Vieira, R.A.; Rocha, R.; Scapim, C.A.; Amaral, A.T., Jr.; Vivas, M. (2016). Selection index based on the relative importance of traits and possibilities in breeding popcorn. Genetics and Molecular Research. 15, 1-10. https://doi.org/10.4238/ gmr.15027719 10 Acta agriculturae Slovenica, 117/1 - 2021 Discrimination of drought tolerance in a worldwide collection of safflower (Carthamus tinctorius L.) genotypes based on selection indices Table supplementary 1: Characteristics of the 100 different genotypes of safflower used in this study Genotype Genotype Geographical Genotype Genotype Geographical Genotype Genotype Geographical code name origin code name origin code name origin G1 A2 Iran (Azerbayejan) G34 Car159 Germany G67 Car64 Slovakia G2 Ac- Stirling Canada G35 Car160 Russia G68 Car67 Germany G3 aC-sunset Canada G36 Car161 Russia G69 Car68 Germany G4 arak 2811 Iran (Markazi) G37 Car169 Hungary G70 Car70 Lybyan G5 C111 Iran(Isfahan) G38 Car175 India (Kusum) G71 Car72 North Korea G6 Car118 india G39 Car181 India G72 Car74 North korea G7 Car 116 india G40 Car188 Poland G73 Car75 North Korea G8 Car 9 Slovaki G41 Car19 Poland G74 Car76 North korea G9 Car100 italy G42 Car190 iran (isfahan) G75 Car77 North korea G10 Car106 Spain G43 Car198 Azerbaijan G76 Car78 Hungary G11 Car114 India G44 Car199 Korean republic G77 Car79 Japan G12 Car117 Sudan (tozi) G45 Car200 unknown G78 Car80 North korea G13 K21 iran (Kordestan) G46 Car201 Sudan G79 Car83 Tajikistan G14 Car124 Pakistan G47 Car210 Spain G80 Car86 Tunisia G15 Car125 Russia G48 Car211 Germany G81 Car87 Romania G16 Car126 Belgium G49 Car214 Poland G82 Car89 Tunisia G17 Car127 Germany G50 Car215 Germany G83 Car94 Spain G18 Car129 Germany G51 Car216 Germany G84 GE62918 Germany G19 Car130 morocco G52 Car217 Germany G85 Gila USA G20 Car131 Paraguay G53 Car218 Germany G86 Hartman USA G21 Car132 Germany G54 Car219 Germany G87 IL111 Iran (aur-oumieh) G22 Car135 Portugal G55 Car221 Germany G88 Isf-14 Iran (Isfahan) G23 Car137 Pakistan G56 Car224 Germany G89 Isf28 Iran(Isfahan) G24 Car138 Poland G57 Car226 Germany G90 K21 Iran (kord-estan) G25 Car146 Egypt G58 Car227 Germany G91 KMS 36 Iran (karaj) G26 Car147 Pakistan G59 Car228 Germany G92 Mex.17-45 Mexico G27 Car148 Pakistan G60 Car230 Germany G93 Mex.7-147 Mexico G28 Car151 india G61 Car24 Morocco G94 Mex.7-38 mexico G29 Car152 iraq G62 Car37 Sudan G95 Mex-13-216 mexico G30 Car155 Russia G63 Car42 Sudan G96 Mex2-138 mexico G31 Car156 Pakistan G64 Car49 Spain G97 Mex22-191 Mexico G32 Car157 morocco G65 Car55 Poland G98 Mex6-97 mexico G33 Car158 Paraguay G66 Car56 Nebraska 8 (USA) G99 PI 301055 Turkey G100 Saffire Canada Acta agriculturae Slovenica, 117/1 - 2021 11