Genome-Wide Association Identifies Candidate Genes for Salt Tolerance in Soybean at Emergence and Seedling Stages

1. Introduction

Soybean (Glycine max [L.] Merr.) is a globally critical food and oilseed crop whose stable production underpins food security and agricultural economies worldwide [1]. Soil salinization severely constrains soybean growth, development, and yield by disrupting cellular homeostasis through ion toxicity, osmotic stress, and oxidative damage, collectively suppressing germination, root development, photosynthetic efficiency, and biomass accumulation [2,3]. Globally, an estimated 950 million hectares are affected by salinity, including roughly 20% of the 230 million hectares of irrigated land, with the affected area expanding at approximately 10% annually [4,5,6,7]. Developing salt-tolerant cultivars is consequently among the most cost-effective strategies to address this challenge [8].

Salt tolerance in soybean is a quantitative trait governed by multiple genes acting through a complex regulatory network [9]. Critically, tolerance is developmentally stage-specific: germplasm exhibiting strong tolerance at germination or seedling establishment does not necessarily perform well at maturity, and tolerance across stages can be largely uncorrelated [10,11,12,13]. Furthermore, salt stress during early development suppresses canopy establishment and vegetative growth, with cascading effects on subsequent stages that ultimately reduce both seed yield and protein content [14,15,16]. Mapping the genetic loci underlying salt tolerance specifically at emergence and seedling stages is therefore a prerequisite for breeding improved cultivars [17].

Early efforts to map soybean salt tolerance relied on biparental populations. Lee et al. [18] used an F_2:5 population from a cross between the salt-tolerant cultivar S-100 and the salt-sensitive cultivar Tokyo, mapping a major seedling-stage QTL to a 3.6 cM interval flanked by SSR markers Sat_091 and Satt237 on chromosome 3. This locus was subsequently validated across multiple studies [19,20,21,22]. In 2014, Qi et al. [23] and Guan et al. [24] independently cloned the underlying gene, Glyma03g32900, from wild and cultivated soybean, designating it GmCHX1 and GmSALT3, respectively; the encoded cation Na⁺/H⁺ transporter limits Na⁺ accumulation in seedling leaves, markedly improving salt tolerance. Do et al. [20] crossed salt-tolerant Fiskeby III with moderately salt-sensitive Williams 82, generating 132 F₂ families, and mapped a QTL on chromosome 13 linked to leaf sodium concentration (LSC). More recently, Li et al. [25] used a recombinant inbred line (RIL) population derived from Williams 82 × PI483460B to map qSalt_Gm18 on chromosome 18, alongside qSalt_Gm03, which co-localizes with GmCHX1 on chromosome 3. To date, 68 soybean salt tolerance QTLs identified from biparental populations have been cataloged in SoyBase (www.soybase.org).

Relative to biparental linkage mapping, genome-wide association studies (GWAS) exploit linkage disequilibrium (LD) within natural populations and offer distinct advantages: broader allelic diversity, finer mapping resolution, and no requirement to construct segregating populations [26,27,28]. High-density SNP-based GWAS has been widely deployed across crops, including rice [29], cowpea [30], rapeseed [31], cotton [32], alfalfa [33], and sesame [34], and has emerged as a cornerstone approach in plant molecular breeding [35]. In soybean, Kan et al. [36] applied GWAS to 191 landraces at the germination stage, identifying one significant SNP on chromosome 9 associated with the germination index ratio and seven SNPs on chromosomes 2, 3, 9, 12, and 13 associated with the germination rate ratio. Dong et al. [37] subsequently identified a major salt tolerance locus controlled by E2, a homolog of Arabidopsis GIGANTEA (GI), using leaf chlorosis as the phenotypic indicator; E2 knockout enhanced tolerance by promoting peroxidase activity and reducing reactive oxygen species (ROS) accumulation under salt stress.

Despite these advances, genetic studies of salt tolerance specifically at the emergence stage remain scarce. In this study, a natural population of 256 soybean accessions was evaluated for salt tolerance at both emergence and seedling stages under controlled salt stress and genotyped with a high-density SNP array. GWAS was then conducted to identify regulatory loci and mine candidate genes, providing a resource for gene cloning and marker-assisted selection in salt-tolerance breeding programs.

2. Materials and Methods

2.1. Materials

A total of 256 soybean accessions were used in this study, comprising 255 domestic and one foreign accession (Table S1). The salt-tolerant cultivar Zhonghuang 39 and the salt-sensitive line NY27-38 served as controls throughout emergence-stage evaluations.

2.2. Salt Stress Treatment and Tolerance Assessment at the Emergence Stage

For each accession, 60 mature, uniformly sized seeds of consistent shape and color were selected. Small pots (7 × 7 × 8 cm) were filled with vermiculite to 2 cm below the rim, sown with 10 seeds each, and topped with vermiculite to the rim. Every 24 pots were arranged in a large container (46 × 32 × 10 cm). For salt treatment, 6 L of 150 mmol/L NaCl solution was added to each large container; after 5 min of saturation, the small pots were transferred to a clean container. Subsequently, 3 L of Reverse Osmosis water (RO water) was replenished every 3 days. Once the vermiculite reached its maximum water-holding capacity, the pots were immediately moved to a fresh container to prevent salt leaching. Control treatments were conducted identically, substituting the NaCl solution with 6 L of water. All treatments were performed in triplicate.

Emergence was scored daily from the appearance of the first seedling, defined as cotyledon protrusion above the vermiculite surface, until NY27-38 displayed clear salt-injury symptoms (severely suppressed emergence and failure of cotyledon expansion) while Zhonghuang 39 remained unaffected, with normal emergence and cotyledon expansion. At this endpoint, the number of fully established seedlings (those with expanded cotyledons and leaves) was recorded. Salt tolerance phenotypes were assessed using an individual-plant classification scheme (Figure 1; Table S2) according to Liu et al. [38], from which the salt tolerance index (SI) was calculated. Emergence-stage salt tolerance grade (STG-SI) was then assigned based on the mean SI across three replicates according to Table S2 (Figure S1). The SI was calculated as:

$$\mathrm{SI}=\sum (\mathrm{I}\times N_i)/10\times 5\hspace{0.33em}$$

where I is the individual-plant category value, Nᵢ is the number of plants in category I, 10 is the number of seeds sown per pot, and 5 is the maximum category value. The salt tolerance coefficient (ST) was defined as the ratio of emerged seedlings under salt stress to those under control conditions.

2.3. Salt Stress Treatment and Tolerance Assessment at the Seedling Stage

Seed preparation and sowing for the seedling-stage assay followed the same protocol as described for the emergence stage. After sowing, 6 L of RO water was added to each large container. After 5 min of saturation, the pots were transferred to a clean container, and 3 L of RO water was replenished every 3 days using the same procedure. Once unifoliate leaves were fully expanded (approximately 10 days after sowing), salt treatment was initiated by adding 3 L of 200 mmol/L NaCl solution to each container. After saturation, pots were moved to clean containers, and this treatment was repeated on days 13 and 16. Five days after the final salt application, leaf salt-injury symptoms were scored in each replicate according to a five-grade scale (Table S3; Figure S2) following Liu [39]: grade 1, highly tolerant; grade 2, tolerant; grade 3, moderately tolerant; grade 4, sensitive; grade 5, highly sensitive. The mean grade across three replicates was calculated to assign the final seedling-stage salt tolerance grade (STG-SS), with means falling between consecutive integers rounded to the higher grade.

2.4. Phenotypic Data Processing and Analysis

Descriptive statistics and normality testing were conducted in SPSS (v25.0, IBM), and one-way ANOVA was performed in R (v4.4.2). The emergence-stage salt tolerance coefficient (ST) and salt tolerance index (SI) were standardized by rank-based inverse normal transformation (INT) following Zachary et al. [40], yielding the transformed traits ST-INT and SI-INT, respectively.

For ordinal traits with small, skewed distributions, binarization is a recognized strategy in GWAS that sharpens contrast between phenotypic extremes and increases power to detect association signals [41], Accordingly, STG-SI and STG-SS were each binarized: accessions scoring below 3 were classified as strongly tolerant (coded 1) and those scoring 3 or above as weakly tolerant (coded 0), producing binary traits STG-SIB and STG-SSB, respectively. Pearson correlation coefficients among ST-INT, SI-INT, STG-SIB, and STG-SSB were computed using the psych package (v2.6.3) in R (v4.4.2) and visualized with the ggpairs function in the GGally package (v2.4.0).

2.5. Genotype Data Quality Control

Genomic DNA was extracted from all 256 accessions and genotyped using the ZDX1 SNP array, co-developed by the Institute of Crop Sciences, Chinese Academy of Agricultural Sciences, and Beijing Compass Biotechnology Co., Ltd. SNP quality control followed Sun et al. [42], with PLINK (v1.9) used to remove SNPs with genotype missing rate > 0.4, or minor allele frequency < 0.05.

2.6. Linkage Disequilibrium Estimation

Linkage disequilibrium (LD) decay was estimated using PopLDdecay (v3.31; BGI Genomics) and visualized in R (v4.4.2) with ggplot2 (v3.4.3). The physical distance at which r² declined to 0.5 (½ max r²) was taken as the LD decay distance for this population.

2.7. Population Structure and Kinship Analysis

K-means clustering was performed using the kmeans function in R (v4.4.2), with the optimal number of subpopulations identified by the elbow method, plotting the sum of squared errors (SSE) for K = 2–10 and selecting the K at which SSE decline flattened. PCA was conducted with the prcomp function in R (v4.4.2). A neighbor-joining (NJ) phylogenetic tree was constructed from pairwise genetic distance matrices using the ape (v5.8-1) and ggtree (v3.16.0) packages. Population structure was additionally estimated with ADMIXTURE across K = 2–8, with the optimal K selected by minimizing cross-validation (CV) error. The final subpopulation assignment integrated results from K-means clustering, PCA, the NJ tree, and ADMIXTURE. A kinship matrix was computed from SNP data using GAPIT (v3.5; http://zzlab.net/GAPIT), and relatedness within and between subpopulations was visualized as a heatmap.

2.8. Genome-Wide Association Analysis (GWAS)

GWAS was conducted on the GAPIT platform (v3.5; http://zzlab.net/GAPIT) using seven models, including GLM, MLM, CMLM, MLMM, SUPER, FarmCPU, and BLINK, with population structure and kinship coefficients included as covariates to control false positives. Significant association loci were declared at a Bonferroni-corrected threshold of −log₁₀(P) ≥ 4, and results were visualized with ggplot2 (v3.4.3) in R (v4.4.2).

2.9. Candidate Gene Identification

LD structure around significant SNPs was examined using LDBlockShow (v1.40) to delineate haplotype blocks and identify candidate gene intervals, guided by the population’s LD decay distance. Expression profiles of candidate genes across soybean tissues were retrieved from the Williams 82 reference genome on Phytozome (https://phytozome.jgi.doe.gov/pz/portal.html).

3. Results

3.1. Phenotypic Evaluation of Salt Tolerance at Emergence and Seedling Stages

At the emergence stage, the mean emergence rate under control conditions was 0.93 (range: 0.4–1.0). Salt stress reduced this significantly to a mean of 0.85 (range: 0–1.0), a difference confirmed by one-way ANOVA (P < 0.0001) (Figure S3).

At the emergence stage, mean ST was 89.32 (range: 33.33–100) and mean SI was 0.62 (range: 0.09–1.00) (Table 1). Both traits deviated markedly from normality: ST was strongly left-skewed (skewness = −1.90) with a leptokurtic distribution (kurtosis = 4.60), while SI exhibited mild left skewness (skewness = −0.50). Rank-based inverse normal transformation was applied to both, and the resulting ST-INT and SI-INT closely approximated normal distributions, with skewness and kurtosis near zero, means near zero, and variance near one (Table 1; Figure S4A, B). STG-SI and STG-SS were both right-skewed (skewness = 0.52 and 0.66, respectively; Table 1) and were therefore binarized, yielding STG-SIB and STG-SSB with strongly tolerant proportions of 60.16% and 62.50%, sufficiently balanced for association analysis (Figure S4C, D). All four transformed traits were carried forward for GWAS.

3.2. Correlation Analysis

Pearson correlation analysis among ST-INT, SI-INT, STG-SIB, and STG-SSB revealed strong positive associations among all three emergence-stage traits (Figure 2): ST-INT vs. SI-INT, r = 0.681; ST-INT vs. STG-SIB, r = 0.520; SI-INT vs. STG-SIB, r = 0.794, all significant. By contrast, the seedling-stage trait STG-SSB showed negligible correlations with STG-SIB (r = 0.062), ST-INT (r = 0.083), and SI-INT (r = 0.106), indicating that salt tolerance at emergence and at the seedling stage are essentially independent phenotypes in this population.

ST-INT is salt tolerance coefficient after inverse normal transformation at emergence stage; SI-INT is salt tolerance index after inverse normal transformation at emergence stage; STG-SIB is salt tolerance index grades after binary conversion at emergence stage; STG-SSB is salt tolerance grades following binary conversion at seedling stage.

3.3. Genotype Data Quality Control

ZDX1 SNP array genotyping across all 256 accessions initially yielded 159,064 SNPs (Figure 3A). After PLINK-based filtering for missing rate and minor allele frequency, 89,361 SNPs were retained for GWAS.

3.4. Linkage Disequilibrium

PopLDdecay analysis showed that r² declined steadily with increasing physical distance, reaching the decay threshold of r² = 0.5 at approximately 50 kb (Figure 3B), which was adopted as the LD decay distance for candidate gene interval definition.

3.5. Population Structure Analysis Based on Clustering, Phylogeny, and Principal Components

Four complementary methods were employed to characterize the genetic structure of this population. K-means clustering with the elbow method revealed a clear inflection point in the within-cluster sum of squares at K = 6, marking this as the most appropriate clustering solution (Figure 4A). PCA was consistent with this result, revealing pronounced genetic stratification in which PC1 and PC2 explained 13.34% and 9.84% of total genetic variation, respectively, with samples forming six discernible clusters in PC1–PC2 space (Figure 4B). NJ phylogenetic analysis independently grouped all accessions into six distinct branches (Figure 4C). ADMIXTURE analysis produced a broadly concordant outcome (Figure 4D,E): cross-validation error declined continuously with increasing K, but the per-step reduction (ΔCV) diminished progressively from K = 2 to K = 6, then rose sharply at K = 7 (ΔCV = 0.0267), pointing to six or seven as the likely range for the true number of subgroups. Integrating all four analyses, K = 6 was selected as the optimal solution for this population, with subpopulations 1 through 6 containing 15, 90, 70, 17, 22, and 35 accessions, respectively; this subpopulation assignment was incorporated into all subsequent GWAS models. The kinship heatmap further confirmed elevated relatedness among individuals within subpopulations alongside clear genetic differentiation between them (Figure 4F), a pattern consistent with within-group homogenization and between-group divergence.

3.6. Genome-Wide Association Study of Salt Tolerance-Related Traits

GWAS was performed on the GAPIT platform for all four transformed traits, ST-INT, SI-INT, STG-SIB, and STG-SSB, using the full SNP dataset. To minimize false positives, only loci detected by at least three of the seven association models were retained. This yielded 60 significant SNPs consolidated into 19 QTLs, with the number of SNPs per QTL ranging from 1 to 20 (Table 2; Figure 5 and Figure 6).

Three QTLs for the emergence-stage trait ST-INT were mapped to chromosomes 4, 7, and 13 (Table 2). The strongest signal was qST-INT-04 (Gm04_46793849) on chromosome 4 (−log₁₀P = 5.7). The chromosome 13 locus qST-INT-13 (Gm13_35700100) was situated approximately 0.87 Mb from GmbZIP132 [43] and 1.39 Mb from GmBIN2 [44]. GmbZIP132 encodes a bZIP transcription factor induced by salt and drought that positively regulates salt tolerance at seed germination [43], while GmBIN2 encodes a GSK3-type kinase similarly induced by both stresses and a positive regulator of stress tolerance [44].

Ten QTLs for the emergence-stage trait SI-INT were identified across chromosomes 1 (four loci), 2, 7, 9, 10, 11, and 14 (Table 2). The strongest signal came from qSI-INT-02 (Gm02_3640916–Gm02_3649811) on chromosome 2, which spanned two SNPs. The chromosome 9 locus qSI-INT-09 (Gm09_5191652) mapped approximately 1.70 Mb from GmPHD6 [45], whose overexpression in soybean hairy roots enhances salt tolerance while knockout increases sensitivity. The chromosome 10 locus qSI-INT-10 (Gm10_3521576–Gm10_3614887) lay approximately 2.54 Mb from GmWRKY54 [46] and 2.14 Mb from GmERF75 [47], both implicated in salt tolerance. The chromosome 14 locus qSI-INT-14 (Gm14_6728116) was located 0.76 Mb from a previously reported marker associated with GsPRX, a gene involved in the oxidative stress response [48].

ST-INT is salt tolerance coefficient after inverse normal transformation at emergence stage; SI-INT is salt tolerance index after inverse normal transformation at emergence stage; STG-SIB is salt tolerance index grades after binary conversion at emergence stage; STG-SSB is salt tolerance grades following binary conversion at seedling stage.

ST-INT is salt tolerance coefficient after inverse normal transformation at emergence stage; SI-INT is salt tolerance index after inverse normal transformation at emergence stage; STG-SIB is salt tolerance index grades after binary conversion at emergence stage; STG-SSB is salt tolerance grades following binary conversion at seedling stage.

Three QTLs for the emergence-stage binary trait STG-SIB were detected on chromosomes 5, 12, and 15 (Table 2), with qSTG-SIB-15 (Gm15_15831968) on chromosome 15 producing the strongest signal (−log₁₀P = 5.74).

Three QTLs for the seedling-stage trait STG-SSB were mapped to chromosomes 3, 10, and 11 (Table 2). The chromosome 3 locus, qSTG-SSB-03 (Gm03_38525925–Gm03_38815289), was the most prominent, spanning a 0.29 Mb interval supported by 20 SNPs, with peak signal at Gm03_38688580 (−log₁₀P = 9.08). This QTL lies approximately 192 kb from Glyma03g32900 (GmSALT3/GmCHX1), the major seedling-stage salt tolerance gene cloned independently by Guan et al. [24] and Qi et al. [23], corroborating earlier reports of a major salt tolerance locus on chromosome 3 [18,21].

3.7. Candidate Gene Analysis

To refine candidate intervals and clarify local haplotype structure, candidate gene identification focused on qSI-INT-10 (Gm10_3521576–Gm10_3614887) for emergence-stage salt tolerance and qSTG-SSB-10 (Gm10_38386826–Gm10_38472859) for seedling-stage salt tolerance, the QTLs with the greatest number of significant SNPs outside the chromosome 3 locus. LDBlockShow analysis resolved seven well-defined LD blocks (Figure S5). For qSI-INT-10, a consistent and significant LD block was detected across the GLM, MLMM, and SUPER models (Figure S5A–C), while for qSTG-SSB-10, a conserved block was identified in the BLINK, CMLM, GLM, and MLM models (Figure S5D–G).

From each stable LD block, the SNP with the highest r² was selected, and genes within a 50 kb window on either side were screened as candidates based on the estimated LD decay distance. Within the emergence-stage QTL qSI-INT-10, 23 candidate genes were identified (Table 3), spanning functions including cell wall biosynthesis and remodeling (e.g., cellulose synthase A4 Glyma.10G039600 and the KATAMARI1 homolog Glyma.10G040700, encoding a xyloglucan galactosyltransferase), transcriptional regulation (e.g., MYB-family transcription factor APL homolog Glyma.10G039700 and auxin response factor 1 Glyma.10G040400), stress response and metabolism (e.g., glutathione S-transferase Glyma.10G040000 and alcohol dehydrogenase 1 Glyma.10G041000), and protein kinase activity and methylation (e.g., Glyma.10G041300 and Glyma.10G040100). For the seedling-stage QTL qSTG-SSB-10, 18 candidate genes were identified (Table 3), enriched for roles in DNA metabolism and repair (e.g., RecQ family helicase Glyma.10G148300, Werner syndrome-like exonuclease Glyma.10G148600, and single-stranded DNA-binding protein Glyma.10G149700) and stress response (e.g., drought-induced protein 19 Glyma.10G149200 and calmodulin-binding protein Glyma.10G148700), along with six uncharacterized and three functionally unknown proteins. Based on functional annotation and established relevance to salt stress, four genes, namely Glyma.10G040000, Glyma.10G148700, Glyma.10G149200, and Glyma.10G149600, were designated high-confidence candidate genes and are discussed in detail below.

4. Discussion

Salt tolerance in soybean is developmentally stage-specific [49], and tolerance at germination generally exceeds that at emergence; timely post-sowing irrigation can reduce surface salinity and partially mitigate this vulnerability [50]. In soils with dry-weight salt content below 1.0%, emergence is reduced even when germination rates remain high across genotypes. This distinction is strikingly illustrated by the cultivar Williams, which maintained a germination rate of 81% at 330 mM NaCl, yet seedling growth plummeted to just 5% under the lower concentration of 220 mM NaCl [51]. Under field conditions, seeds may germinate successfully in saline soil but fail to penetrate a salt-hardened surface crust. This demonstrates that soybean seeds can survive saline conditions during germination, while the same stress can be lethal at emergence, underscoring the need to evaluate salt tolerance, map relevant loci, and identify candidate genes specifically at the emergence stage.

Relatively few soybean salt-tolerance genes have been cloned through forward genetics to date, including the seedling-stage genes GmSALT3/GmCHX1 [23,24] and GsERD15B [48], and the germination-stage gene GmCDF1 [52]. The present GWAS yielded four candidate genes, one emergence-stage and three seedling-stage. The emergence-stage candidate Glyma.10G040000, located within qSI-INT-10, encodes a glutathione S-transferase (GST). GSTs are well-established contributors to plant salt tolerance, acting primarily by maintaining ROS homeostasis and thereby limiting salt-induced oxidative damage; transgenic overexpression of GST genes consistently enhances salt tolerance [53]. GST expression is further modulated by MYB and WRKY transcription factors, embedding these enzymes within broader stress-regulatory networks [54]. In soybean specifically, the tau-class member GmGSTU23 enhances salt tolerance through elevated glutathione transferase activity [55], suggesting that Glyma.10G040000 may confer tolerance through analogous mechanisms centered on ROS detoxification.

The three seedling-stage candidate genes all map to qSTG-SSB-10. Glyma.10G148700 encodes a calmodulin-binding protein, a class of regulators that decode salt-induced cytosolic Ca²⁺ transients and translate them into adaptive physiological responses. In rice, the calmodulin-binding protein OsMSR2 is strongly induced by salt stress, and its overexpression in Arabidopsis enhances tolerance by modulating ABA signaling and downstream stress-response gene expression [56]. Conversely, the Medicago truncatula ortholog MtCML40 acts as a negative regulator: its overexpression increases salt sensitivity, accompanied by Na⁺ accumulation in shoots and suppression of the sodium efflux genes MtHKT1;1 and MtHKT1;2 [57]. The contrasting roles of these orthologs highlight calmodulin-binding proteins as pivotal, if context-dependent, regulators of the salt-stress response, and position Glyma.10G148700 as a strong candidate for further functional characterization.

Glyma.10G149200 encodes drought-induced protein 19 (Di19), a family with documented roles in plant salt tolerance. In soybean, GmDi19-5 is a negative regulator that interacts with the E3 ubiquitin ligase GmPUB21 and is degraded in an ABA-dependent manner, enabling fine-tuned modulation of stress responses [58]. In maize, ZmDi19-1 enhances salt tolerance by activating downstream stress-responsive genes and boosting antioxidant capacity [59], while in cotton, GhDi19-3 and GhDi19-4 improve salt tolerance through Ca²⁺ and ABA signaling and concurrent ROS scavenging [60]. Collectively, Di19 proteins appear to function as central signaling nodes integrating multiple stress pathways, making Glyma.10G149200 a compelling candidate for salt tolerance.

Glyma.10G149600 encodes a PP2C family protein. In Arabidopsis, PP2Cs are core negative regulators of ABA signaling and can directly interact with and inhibit the plasma membrane Na⁺/H⁺ antiporter SOS1, a key sodium efflux transporter [61]. In peanut, PP2C genes including AhPP2C45 and AhPP2C134 are significantly upregulated under salt stress [62]. By directly regulating sodium transport and compartmentalization, PP2Cs serve as indispensable nodes in saline-environment adaptation, supporting a role for Glyma.10G149600 in soybean salt tolerance.

Beyond gene discovery, translating genomic findings into practical breeding requires broader consideration of how and when salt stress is evaluated. Unlike many abiotic stresses, salt-alkali stress persists throughout most or all of the crop growth cycle [63]. Current soybean salt-tolerance research is disproportionately focused on the seedling stage, under the assumption that early vegetative vigor reliably predicts agronomic performance, an assumption that does not always hold, since excessive salinity can redirect photosynthate from growth to stress tolerance. Indeed, some salt-tolerant genotypes show early vigor comparable to sensitive lines yet significantly outperform them under field conditions [50], highlighting the indispensable role of multi-stage field evaluation. Furthermore, most laboratory assays use NaCl as the sole treatment, whereas salinized soils worldwide vary widely in pH and ionic composition [64]. Salinization and alkalization frequently co-occur, yet their combined effects on plant growth are not simply additive, the severity ranking from greatest to least is saline-alkaline stress, alkaline stress, then salt stress alone, and mixed saline-alkaline injury substantially exceeds the sum of the individual stresses [65,66]. As research on mixed saline-alkaline tolerance remains sparse, future efforts should extend beyond single-salt or single-alkali treatments to encompass the complex stress combinations encountered in the field.

5. Conclusions

GWAS combining 200K SNP array data from 256 soybean germplasm accessions with salt-tolerance phenotypes at emergence and seedling stages identified 60 significant SNPs consolidated into 19 QTLs, with 13 associated with emergence-stage and three with seedling-stage salt tolerance. Known stress-tolerance genes were located near five of these QTLs; notably, the seedling-stage QTL qSTG-SSB-03 lies just 192 kb from the major salt-tolerance gene GmSALT3. Four high-confidence candidate genes were prioritized: the emergence-stage gene Glyma.10G040000, encoding a glutathione S-transferase, and the seedling-stage genes Glyma.10G148700, Glyma.10G149200, and Glyma.10G149600, encoding a calmodulin-binding protein, drought-induced protein 19, and a PP2C family protein, respectively. Functional validation of these candidates will be an important priority for future work.