The experimental setup for cultivating safflower genotypes utilized a soil mixture consisting of sand and soil in a 1:2 ratio, which was used to fill the pots. A comprehensive soil analysis was conducted, with micro nutrients assessed using atomic absorption and macro nutrients evaluated via flame photometry. Each pot contained 780 g of soil that had been passed through a 4 mm sieve. The pots were placed in a glasshouse environment, maintaining a temperature range of 26-28 °C and 46% humidity. The experiment was conducted under a light cycle of 16 hours followed by 8 hours of darkness. Initially, three seeds of each safflower genotype were planted in each pot, with subsequent thinning after germination to retain only one seedling per pot for data collection. The plants were watered regularly with tap water, adjusting the frequency based on the specific requirements of each genotype.

The study utilized 94 safflower genotypes sourced from 26 countries, procured from the United States Department of Agriculture (USDA) (Supplementary Table 1). The experiment, initiated on February 10th, 2024, employed a two-factorial randomized complete block design with three replications and four treatments. The experiment involved three plants per genotype for data collection.

The four NaCl treatments included: T1 (control with no NaCl), T2 (80 mM NaCl), T3 (160 mM NaCl), and T4 (250 mM NaCl). These treatments were selected based on established literature for safflower salt stress studies and preliminary experiments conducted in our lab (Thoday-Kennedy et al., 2021). These concentrations span control to severe stress levels commonly applied in safflower phenotyping, allowing discrimination of tolerant genotypes while avoiding complete lethality. 250 mM concentration simulates extreme field-relevant salinity, within safflower’s moderate tolerance range but sufficient to identify resilient genotypes without universal mortality, consistent with hydroponic and soil-based assays (Thoday-Kennedy et al., 2021). The NaCl applications in T2, T3, and T4 were administered between 16 and 35 days after sowing, spread over several days to prevent salt shock. At 35 days post-sowing, the safflower genotypes were harvested, and key morphological traits were recorded. For the association analysis, the mean values of each trait across all treatments were utilized.

Plant height (PH) was determined by measuring from the stem base to the shoot apex in centimeters with the help of a scale. The number of leaves (NL) were counted on each selected plant manually. The same plant was then weighed to record its biological yield (BY) in grams. For root length (RL) measurement, each plant was carefully uprooted and washed with tap water. RL was measured in centimeters with the help of a scale. Fresh root (FRW) and shoot (FSW) weights were recorded in grams using a digital balance. The fresh shoots and roots of the selected plants were oven-dried and then weighed to measure the dry shoot (DSW) and root (DRW) weights.

DNA was extracted from two-week-old safflower seedlings using two methods: the CTAB protocol as described by Doyle and Doyle in 1990 (Doyle and Doyle, 1990), and a specific protocol recommended by Diversity Arrays Technology. The latter protocol is available on the Diversity Arrays Technology website (https://www.diversityarrays.com/services/laboratory-services/).

The soil composition comprised a blend of sand and soil, mixed in a proportion of one part sand to two parts soil. The concentrations of micro and macronutrients present in this mixture are detailed in Table 1.

The analysis of variance (ANOVA) results indicated statistically significant variations among safflower genotypes, NaCl treatments (T), and their interaction (G * T) for the majority of traits examined. However, BY and FSW were exceptions, showing no significant differences in these analyses (Table 2).

The varying levels of NaCl had a significant impact on all the studied traits across the treatments (Table 3). For instance, PH decreased from 5.64 cm in T1 to 4.96, 4.36, and 3.82 cm in T2, T3, and T4, respectively. FRW declined from 0.40 g in T1 to 0.35, 0.28, and 0.22 g in T2, T3, and T4, respectively. FSW dropped from 0.49 g in T1 to 0.42, 0.37, and 0.30 g in T2, T3, and T4, respectively. NL decreased from 5.00 in T1 to 4.60, 4.20, and 4.05 in T2, T3, and T4, respectively. RL reduced from 4.62 cm in T1 to 4.15, 3.64, and 3.15 cm in T2, T3, and T4, respectively. Likewise, BY declined from 0.88 g in T1 to 0.77, 0.67 and 0.51 g in T2, T3, and T4, respectively. DSW decreased from 0.21 g in T1 to 0.16, 0.12, and 0.07 g in T2, T3, and T4, respectively, while DRW dropped from 0.17 g in T1 to 0.13, 0.08, and 0.04 g in T2, T3, and T4 (Table 3). Varying NaCl levels significantly affected the average performance of all the studied traits. Notable differences were observed among treatments for all traits, except for NL, where no differences were detected between T3 and T4 (Figure 1). The values of broad sense heritability ranged from 17% to 97% across the evaluated traits (Table 4).

The correlation analysis revealed significantly positive relationships between most of the studied traits across all treatments (Figure 2). In T1, all traits showed significantly positive correlations with each other. In T2, most traits exhibited significantly positive correlations, with the exception of FSW, which negatively correlated with all the studied traits. In T3, positive correlations were observed among most traits, except for NL and DRW, which negatively correlated with BY and DSW, respectively. Similarly, in T4, all traits were in positive correlation among each other (Supplementary Table 2). The correlation analysis using mean data across all treatments also demonstrated significantly positive correlations among all the traits examined except FSW and BY.

The first two principal components, each having eigenvalues of 1 or greater, explained a considerable portion of the total variation (40.8%) in PCA. The first principal component (PC1) accounted for 31.4% of the overall variance, while the second principal component (PC2) explained an additional 9.4% variance (Figure 3). Based on their distribution in the PCA biplot quadrants, all genotypes were categorized into three distinct groups, labeled as 1, 2, and 3.

A rigorous selection criterion identified top-performing genotypes (G-50, G-58, and G-94) from the safflower panel using genotypic data and trait mean data under all treatments (Table 5). The STI indicated moderate to high tolerance in the majority of genotypes across all studied traits (Supplementary Table 3). The genotypes ranking was further confirmed using online tool iPASTIC (Supplementary Table 4). Genotypes demonstrating the best mean performance and highest salt tolerance could be valuable for developing salt-tolerant cultivars.

In this study, 12,232 highly informative DArTseq markers - filtered from GBS data generated by Diversity Arrays Technology (DArT) for 94 safflower genotypes - were used to perform GWAS on eight morphological traits at the seedling stage using the MLM (Q + K) model. Phenotypic data from these traits, collected under four NaCl treatments under control conditions, enabled identification of environmentally stable and treatment-specific MTAs. The SNP-GWAS analysis identified 322 significant MTA loci across the eight morphological traits under all four treatments. Specifically, 34, 25, 44, and 48 MTAs were detected for BY, DRW, DSW, and FRW, respectively, while FSW, NL, PH, and RL showed 46, 60, 47, and 18 MTAs (Supplementary Table 5; Figure 4). A total of 17, 21, 41, 12, 24, and 42 MTAs were detected on chromosomes 1–6, while chromosomes 7–12 harbored 25, 44, 19, 17, 42, and 18 MTAs, respectively (Figure 5). Overall, these MTAs hold promise for improving morphological traits in safflower.

To investigate molecular mechanisms of salinity tolerance in safflower, we identified loci associated with GWAS markers and their Arabidopsis orthologues via BLAST searches (Table 6). Marker DArT-100005876 (linked to BY in T1) yielded six orthologues, while others corresponded to single orthologues. PPI networks were constructed for each orthologue across treatments (Figures 6–9; Supplementary Table 6). In T1, PHOSPHOLIPASE A I (PPI 1 of marker DArT-100005876) interacted with 11 proteins, ADENYLYL-SULFATE KINASE 4 (PPI 2) with 21, GLUCOSAMINE INOSITOLPHOSPHORYLCERAMIDE TRANSFERASE 1 (PPI 3) and TPLATE (PPI 4) each with 21, and RIBOSOMAL PROTEIN UL13M (PPI 5) with 11. In T2, networks centered on SUCROSE-PHOSPHATE SYNTHASE 2 (PPI 1, marker DArT-38084188; FRW; 11 proteins), AT1G33770 protein kinase (PPI 2, marker DArT-38083288; FRW; 31 proteins), and FAR1-RELATED SEQUENCE 3 (PPI 3, marker DArT-45483426; RL; 21 proteins). In T3, AT5G01610 (PPI 1, marker DArT-15670616; DSW; 21 proteins) and DETOXIFICATION 50 (PPI 2, marker DArT-100020623; DSW; 11 proteins). In T4, RALFL1 (marker DArT-22763645 associated with DSW) included 21 proteins. These PPI networks provide critical insights into the potential molecular interactions contributing to salinity tolerance in safflower, highlighting key candidate genes and their associated pathways.

To gain a deeper understanding of the functional relevance of the protein clusters identified in the PPI network analysis, we conducted GO enrichment analysis based on their biological functions and metabolic processes (Supplementary Figures 1–4). GO enrichment of PPI clusters revealed treatment-specific pathways: T1 featured lipid metabolism/jasmonic acid biosynthesis (PPI 1), sulfate assimilation and metabolism (PPI 2), sphingolipid biosynthesis (PPI 3), ribosome biogenesis and rRNA processing (PPI 4), and translation (PPI 5) (Supplementary Figure 1). T2 showed sucrose metabolism (PPI 1), pectin catabolism (PPI 2), and nitrate import/root regulation (PPI 3) (Supplementary Figure 2). T3 involved ammonia assimilation and carboxylic acid transport (PPI 1) and ABA transport (PPI 2) (Supplementary Figure 3). T4 highlighted protein autophosphorylation/brassinosteroid responses (Supplementary Figure 4). These networks illuminate interconnected pathways in safflower salinity responses.

Integrating GWAS signals with transcriptome data enhances candidate gene prioritization for salinity tolerance. Co-localization identified 11 highly expressed DEGs across eight genomic regions, categorized by z-score normalized expression into two clusters (four and seven genes; Figure 7). Notable upregulated genes under salt stress included PLA1 (vacuole biogenesis/generative organ development), SPS2 (sucrose accumulation in response to stresses like salinity, supporting metabolism), and DTX50 (transmembrane protein in SOS pathway essential for salt stress). This convergence strengthens their candidacy for functional validation in safflower breeding.

ANOVA revealed significant differences among genotypes, treatments, and their interaction (G * T) for most studied traits, with the exceptions of BY and FSW (Table 2). The presence of significant differences among genotypes for the studied traits highlights the availability of genetic variation for safflower improvement, which corroborates previous findings (Ali et al., 2025). Among the treatments, T2 (80 mM) was identified as the salt dose causing the least reduction in morphological traits, while T4 showed the maximum decrease across all studied traits.

The germplasm’s overall performance in this study was significantly impacted by varying NaCl treatments (Supplementary Figure 5). For instance, PH was decreasing 12.06%-32.27% in T2-T4, compared with T1. FSW was reducing 14.29%-38.78% in T2-T4, compared with T1. NL was diminishing 8.00%-19.00% in T2-T4, compared with T1. FRW was declining 12.50%-45.00% in T2-T4, compared with T1. RL was decreasing 10.17%-31.82% in T2-T4, compared with T1. BY was reducing 12.50%-42.05% in T2-T4, compared with T1. DSW was diminishing 23.81%-66.67% in T2-T4, compared with T1. DRW was decreasing 23.53%-76.47% in T2-T4, compared with T1. These findings align with previous research on safflower and other crops under salt stress conditions (Gengmao et al., 2015; Uzair et al., 2022). The results also indicate that safflower exhibits moderate salt tolerance, highlighting the need for further research to identify genotypes and genes that could contribute to the development of salt-tolerant cultivars.

Previously, moderate salt tolerance capability in safflower, with 50% reduction in biomass at 125 mM NaCl (12.5 dS/m) has been reported. Growth is severely impacted at higher NaCl concentrations of 250 mM (25 dS/m) and 350 mM (35 dS/m) (Gengmao et al., 2015; Singh et al., 2013; Kaya et al., 2003, Kaya et al., 2019; Ghazizade et al., 2012). In another study on safflower, genotypes were evaluated for salt tolerance using 150 mM NaCl at the vegetative stage, with total plant biomass and photosynthetic attributes serving as selection criteria (Siddiqi et al., 2009). Genotypic variation is more pronounced under higher NaCl treatments, making moderate to severe stress conditions optimal for identifying stress tolerance in diverse germplasm.

The present study reported high heritability for most of the traits examined, except for BY and FSW, which showed low heritability estimates. Heritability values are important indicators of the potential effectiveness of selection in plant breeding. A high heritability value suggests that most of the observed variation is due to genetic differences among the plants, making selection more efficient. In contrast, low heritability indicates that the variation is largely attributable to environmental factors rather than genetics, meaning that the genotype is heavily influenced by the testing environment (Limbongan et al., 2024).

Correlation analysis reveals valuable relationships between traits, aiding crop improvement (Ali et al., 2020b). Under salt stress, significantly positive correlations exist between shoot length and root length, as well as between seedling fresh weight and both root and shoot lengths. Additionally, root fresh weight correlates positively with seedling length, root length, and seedling fresh weight (Kayaçetin, 2020; Gürsoy, 2023). These findings collectively reinforce the interconnectedness of plant growth parameters in response to salt stress. Furthermore, traits that show significant associations with one another may be useful for selecting promising genotypes.

PCA simplifies complex data into manageable groups, revealing underlying relationships. It helps select optimal genotypes by considering multiple traits (Ali et al., 2024a). In our study, the first two principal components explained 40.8% of the variation. Genotypes were grouped into three clusters based on traits like PH, FSW, RL, and BY (Supplementary Table 7). Group 3 had the highest PH (4.94 cm) and RL (4.08 cm), while Group 1 had the highest FSW (0.53 g) and BY (0.92 g). The best-performing genotypes identified were G-50 and G-94 from Group 1, and G-58 from Group 2.

Phenotypic data along with population mean and percent increase were employed to identify superior genotypes. Three genotypes, G-50, G-58, and G-94 were emerged as the most favorable ideotypes, demonstrating positive genetic improvements for all traits (Table 5). These genotypes, possessing more favorable alleles, can be utilized in safflower breeding programs to enhance traits associated with salt stress tolerance.

The STI calculation revealed that a majority of genotypes exhibited moderate to high tolerance against salt stress across all studied traits, corroborating previous findings. According to Rahman et al., 2017, higher STI values during the early seedling stage indicate greater tolerance to abiotic stress. Furthermore, the best performing safflower genotypes (G-50, G-58, and G-94) observed high tolerance (> than 0.65) against salt stress in all treatments (Supplementary Table 3) (Tao et al., 2021). The same genotypes were also ranked as tolerant using the iPASTIC online tool (Supplementary Table 4). It is therefore suggested to use the identified genotypes as parental lines for the development of salt-tolerant safflower cultivars.

The identification of loci influencing key plant traits is crucial for marker-assisted breeding aimed at enhancing crop productivity under changing climate conditions. Despite its importance, there are few reports identifying markers or loci associated with agronomic traits in safflower (Ali et al., 2024). Moreover, even fewer studies have reported MTAs for safflower traits under salinity stress. The number of significantly associated markers increased across treatments, with 18 markers showing associations with multiple traits or demonstrating consistency under varying salt stress conditions. Among these consistent markers, DSW was associated with 14 markers, FSW was associated with 11 markers, and DRW was associated with eight markers. Additionally, FSW was associated with three markers. Given these findings, it is recommended that the identified markers be validated and converted into Kompetitive Allele Specific PCR (KASP) assays to accelerate safflower breeding under saline stress conditions. Quantitative traits are influenced by both genetic and environmental factors, although specific details remain to be fully elucidated (Yang et al., 2021). Similar findings have been reported in crops like wheat, where 11 stable markers were identified for breeding under stress conditions (Safdar et al., 2020). This suggests a strong potential for the use of safflower markers in enhancing salt tolerance via marker-assisted breeding.

The PPI 1 in T1 includes PLA1 enzymes, which catalyze the hydrolysis of acyl groups from phospholipids at the sn-1 position. They are crucial for generative organ development and vacuole biogenesis, including controlling polar vacuole location in zygotes for asymmetrical cell division (Figure 6) (Takáč et al., 2019; Kimata et al., 2019). PPI 2 in T1 included APK4, which supplies activated sulfate for sulfating secondary metabolites like glucosinolates (Halkier and Gershenzon, 2006). Sulfated metabolites contribute to plant defense against biotic/abiotic stresses, while small sulfated peptides (e.g., phytosulfokines, PSY1, RGF) regulate growth. These compounds underscore sulfation’s dual role in stress adaptation (e.g., salt tolerance) and developmental processes (Figure 6) (Matsuzaki et al., 2010). PPI 3 in T1 included GINT1, which mediates GIPC glycosylation by adding N-acetylglucosamine. In Arabidopsis, GINT1 inhibition reduces salt sensitivity, with mutants showing higher germination rates under salt stress (Figure 6) (Gigli-Bisceglia and Testerink, 2021). PPI 4 in T1 included TPLATE, involved in cytokinesis and localized to the cell plate. The TPLATE complex is a key adaptor for Clathrin-mediated endocytosis in plants, crucial for growth and physiological processes like nutrient uptake and pathogen defense (Figure 6) (Narasimhan et al., 2020). PPI 5 in T1 includes UL13M, part of the ribosomal protein L13 family. Upregulation of these proteins may enhance translation or ribosome assembly. Overexpressing a homolog in E. coli improved salt and freezing tolerance (Figure 6) (Tanaka et al., 2001).

PPI 1 in T2 includes SPP2, crucial for plant metabolism. Sucrose often accumulates in response to environmental stresses like cold, salinity, and drought (Figure 8) (Lunn and Ap Rees, 1990). PPI 2 involves FRF proteins from the FRS-FRF family, which regulate developmental processes and stress responses (Figure 8) (Jafari et al., 2025). PPI 3 includes AT1G33770, a potential protein kinase related to stress responses like salinity (Figure 8) (Majeed et al., 2023).

PPI 1 in T3 included AT5G01610, a hypothetical protein with a DUF538 domain, whose function remains unknown (Figure 9) (Ijaq et al., 2015). PPI 2 in T3 included DTX50, a transmembrane protein crucial for salt stress response via the SOS pathway, enhancing tolerance to drought, salt, and cold (Figure 9) (Lu et al., 2019). PPI in T4 included RAF1, which activates SnRK2 kinases for salt stress tolerance and ABA signaling (Figure 10) (Fàbregas et al., 2020).

Our analysis revealed that PLA1, SPS2, and DTX50 formed a distinct cluster and were significantly upregulated in the roots following salinity treatment. In contrast, most of the other genes were predominantly downregulated and clustered separately. Notably, AT5G01610 was grouped with PLA1, SPS2, and DTX50 despite exhibiting a slight downregulation in the roots under salinity stress. This clustering pattern suggests that PLA1, SPS2, and DTX50 may play a coordinated role in the root’s adaptive response to salinity, potentially contributing to stress tolerance mechanisms (Figure 7). The association of AT5G01610 with this cluster, despite its mild downregulation, implies that it may share functional or regulatory interactions with these genes, warranting further investigation into its potential role in salinity adaptation.