Genome-Wide Association Study Reveals Key Genetic Loci Controlling Oil Content in Soybean Seeds

Abstract

Seed oil represents a key trait in soybeans, which holds substantial economic significance, contributing to roughly 60% of global oilseed production. This research employed genome-wide association mapping to identify genetic loci associated with oil content in soybean seeds. A panel comprising 341 soybean accessions, primarily sourced from Northeast China, was assessed for seed oil content at Heilongjiang Province in three replications over two growing seasons (2021 and 2023) and underwent genotyping via whole-genome resequencing, resulting in 1,048,576 high-quality SNP markers. Phenotypic analysis indicated notable variation in oil content, ranging from 11.00% to 21.77%, with an average increase of 1.73% to 2.28% across all growing regions between 2021 and 2023. A genome-wide association study (GWAS) analysis revealed 119 significant single-nucleotide polymorphism (SNP) loci associated with oil content, with a prominent cluster of 77 SNPs located on chromosome 8. Candidate gene analysis identified four key genes potentially implicated in oil content regulation, selected based on proximity to significant SNPs (≤10 kb) and functional annotation related to lipid metabolism and signal transduction. Notably, Glyma.08G123500, encoding a receptor-like kinase involved in signal transduction, contained multiple significant SNPs with PROVEAN scores ranging from deleterious (−1.633) to neutral (0.933), indicating complex functional impacts on protein function. Additional candidate genes include Glyma.08G110000 (hydroxycinnamoyl-CoA transferase), Glyma.08G117400 (PPR repeat protein), and Glyma.08G117600 (WD40 repeat protein), each showing distinct expression patterns and functional roles. Some SNP clusters were associated with increased oil content, while others correlated with decreased oil content, indicating complex genetic regulation of this trait. The findings provide molecular markers with potential for marker-assisted selection (MAS) in breeding programs aimed at increasing soybean oil content and enhancing our understanding of the genetic architecture governing this critical agricultural trait.

1. Introduction

Soybean (Glycine max) accounts for 60% of global oilseed production and provides over 25% of the world’s protein for humans and animals [1,2]. Originating from wild soybean (Glycine soja) in central China about 5000 years ago, it has since become widespread worldwide [3,4]. Modern soybean seeds generally contain 17% oil, 35% protein (both essential and non-essential amino acids), 31% carbohydrates (soluble and insoluble), 13% moisture, and 4% ash [5,6]. Depending on variety and growing conditions, oil content ranges from 8.3% to 27.9%, while protein concentration varies from 34.1% to 56.8% [3]. Soybean oil primarily consists of fatty acids (FAs), triacylglycerols (TAGs), and tocopherols, with five main fatty acids determining oil quality: stearic acid (C18:0), oleic acid (C18:1), linoleic acid (C18:2), linolenic acid (C18:3), and palmitic acid (C16:0) [7,8,9]. As a major cooking oil with health benefits, its synthesis pathways remain a significant focus of research [10,11].

High-throughput sequencing and genome-wide association studies (GWASs) have accelerated the identification of quantitative trait loci (QTLs) and genes linked to agronomic traits across various crops [12,13]. The oil content in soybeans is quantitatively inherited and regulated by multiple genes [14].The GWAS methodology employs natural populations with extensive recombination events, thereby improving the accuracy of phenotype association compared to biparental populations [15,16,17]. This approach has been used to identify genetic regions associated with numerous plant traits including oil content, protein content, yield, quality, and stress responses [18,19,20]. To date, genomic analyses have identified over 300 QTLs associated with soybean seed oil content, building upon previous research documented in the SoyBase database (https://www.soybase.org).

Recent molecular breeding efforts have utilized GWASs to identify oil synthesis genes and develop useful markers. Notable studies include analysis genotyping of 320 soybean accessions revealing 29 oil-related QTLs, with 24 being novel discoveries; and GWASs on 278 soybean accessions identifying three significant QTLs, with SNP ss715637321 on chromosome 20 coinciding with previously documented large-effect oil QTL loci [21]. Additional studies have identified the GmST05 gene affecting seed size through a GWAS of over 1800 soybean samples, discovered a SEIPIN homologous gene on chromosome 9 regulating fatty acid synthesis from analysis of 547 soybean samples, and identified 5 oil-related QTLs through QTL mapping in a recombinant inbred line population [22,23,24,25,26].

Triacylglycerol (TAG) synthesis, the predominant form of seed oil accumulation, is primarily governed by the diacylglycerol acyltransferase (DGAT) gene, with overexpression studies showing enhanced oil accumulation [27,28]. Several genes influence soybean oil content and synthesis, including GmOLEO1 (enhancing oil accumulation through TAG synthesis), GmSWEET39 (positively influencing total oil content), and transcription factors like WRI1 (regulating key genes in glycolysis and fatty acid synthesis) [9,29,30,31,32]. Additional transcription factors such as LEC1, LEC2, MYB, ABI3, and bZIP also play essential roles in seed oil accumulation [33,34,35,36]. The regulation of oil biosynthesis in soybean involves complex transcriptional and epigenetic mechanisms that coordinate fatty acid synthesis during seed development. The WRINKLED1 (WRI1) transcription factor, a member of the AP2/ERF family, serves as a master regulator by directly activating genes encoding acetyl-CoA carboxylase, fatty acid synthase subunits, and other key enzymes in the fatty acid biosynthesis pathway [37,38]. In soybean, multiple WRI1 homologs exhibit tissue-specific expression patterns during seed filling, with GmWRI1a and GmWRI1b showing distinct temporal regulation [39,40]. Additionally, MYB transcription factors, including MYB89 and MYB96, modulate fatty acid desaturase gene expression and coordinate oil accumulation with seed maturation processes [41,42]. Epigenetic regulation through DNA methylation and histone modifications further fine-tunes the expression of oil biosynthesis genes, with dynamic methylation changes observed in gene promoter regions during seed development [43,44]. These multilayered regulatory networks suggest that genetic variants affecting transcriptional or epigenetic control mechanisms significantly influence oil content variation, highlighting the importance of comprehensive functional annotation in association studies.

Despite extensive research on seed-oil-based genome-wide association studies, much work remains to fully understand the molecular basis of seed oil content and composition. In our study, a natural population of 341 germplasm resources was evaluated for soybean oil content over two years using phenotypic data and whole-genome resequencing. The genome-wide association study identified significant SNP regions linked to soybean oil content, while subsequent QTL mapping helped pinpoint candidate genes. Analyzing these genes advances our understanding of the genetic factors affecting soybean oil content. These findings offer valuable support for developing high-oil soybean varieties through biotechnological breeding methods, ultimately boosting soybean oil production and improving molecular design breeding strategies.

2. Materials and Methods

2.1. Experimental Materials and Field Trials

The primary experimental materials for this study consisted of an association panel comprising 341 soybean germplasms selected based on their wide range of oil content phenotypes, distant genetic relationships, and significant geographical differences, making them well-suited for studying the genetic basis of oil content variation. A total of 340 individuals were identified from China, with 336 originating from Northeast China and 4 from Northwest China. Additionally, 1 individual was identified from the United States. A total of 153 materials, accounting for 44.87%, originated from Jilin Province. Heilongjiang Province contributed 126 materials, representing 36.95%. Liaoning Province provided 47 materials, which is 13.78% of the total. The remaining materials were sourced from various other regions in lesser amounts.

All 341 materials were planted in both 2021 and 2023 growing seasons from mid-May to early October at the National Modern Agricultural Experimental Farm of Heilongjiang Academy of Agricultural Sciences, located in Minzhu Town, Daowai District, Harbin City, Heilongjiang Province. Climate conditions of the two years are presented in Table S1. The climatic conditions differed substantially between the 2021 and 2023 growing seasons, which likely contributed to the observed variations in soybean oil content. In 2021, the growing season was characterized by relatively moderate temperatures and variable precipitation patterns. Monthly mean temperatures ranged from 7.6 °C (October) to 25.6 °C (July), with total precipitation of 357.3 mm distributed unevenly across months. The season experienced lower relative humidity (45–62%) and moderate sunshine hours (180–265 h monthly). Notably, August showed high precipitation (106.4 mm) with 15 rainy days, potentially affecting pod filling stages. In contrast, 2023 exhibited warmer temperatures and higher precipitation, particularly during critical reproductive phases. Mean temperatures were consistently higher, ranging from 8.5 °C (October) to 23.0 °C (July), representing approximately 1–2 °C increases compared to 2021. Total precipitation reached 532.7 mm, with July showing exceptionally high rainfall (221.6 mm, 22 rainy days) during flowering and early pod development. The 2023 season also featured higher relative humidity (52–75%) and increased wind speeds. These contrasting environmental conditions—cooler, drier conditions in 2021 versus warmer, wetter conditions in 2023—provided an excellent natural experiment for evaluating the stability of genetic associations across diverse climatic environments, strengthening the robustness of the identified GWAS associations.

A fully randomized block design was utilized, consisting of single-row plots with a row length of 3 m, row spacing of 65 cm, and in-row plant spacing of 5 cm. Each germplasm was replicated three times under various conditions in each growing season. Field management adhered to prescribed protocols. Upon maturation, seeds were harvested and naturally sun-dried for phenotyping of oil content.

2.2. Determination of Soybean Seed Oil Content

After the field harvest, the oil content of dried soybean seeds gathered during the 2021 and 2023 growing seasons was quantitatively evaluated using an NIRS DS 2500 (FOSS, Denmark) near-infrared analyzer. This non-destructive spectroscopic method enables swift and precise assessment of seed composition parameters by analyzing specific wavelength absorption patterns. Each biological sample underwent triplicate measurements to guarantee analytical precision and reliability. The calculated arithmetic mean was then used as the definitive phenotypic data point for all subsequent statistical analyses. To validate the accuracy of NIRS measurements, a subset of samples (n = 30) was cross-validated using the traditional Soxhlet extraction method, confirming the reliability of the spectroscopic approach for oil content determination.

2.3. Genome-Wide Association Analysis

The 341 soybean varieties were sequenced using the Hi-Seq 2000 high-throughput sequencing platform, with an average sequencing depth of 10×. The sequencing data were processed using the Genome Analysis Toolkit (GATK) [45], and the filtered data were aligned to the soybean reference genome (v. Wm82.a2) using BWA software (Version: 0.6.1-r104) [46] with default parameters. SNP markers were collected and sorted using GATK, then recalibrated using a Gaussian mixture model to remove outliers. SNPs with a missing rate ≥ 10% and minor allele frequency ≤ 5% were filtered out using PLINK [47]. Linkage disequilibrium (LD) analysis was performed using PopLDdecay software (v3.42) [48], with the LD coefficient R² as the metric. The LD values were calculated using the LD synthesis method in the R package SNPRelate (R version 4.4.3), and highly linked SNPs were pruned. The LD decay plot was adjusted using the LOWESS tool in R, with a smoothing parameter of 0.01. Population structure was analyzed using fastSTRUCTURE v2.3.1 software [49] and principal component analysis (PCA) in the R package SNPRelate (R version 4.4.3). Kinship analysis was performed using GAPIT software (v3) [50]. Significant SNP loci were identified based on 341 varieties and 1,048,576 SNP markers, using the MLM (mixed linear model) in GAPIT software (v3) for genome-wide association analysis. The genome-wide significance threshold was set at p < 10⁻⁶, which was determined using Bonferroni correction for multiple testing to control the family-wise error rate across all tested SNP markers, ensuring stringent control of false positive associations.

2.4. Identification and Validation of Candidate Genes

In this study, the linkage disequilibrium (LD) decay distance was considered, and genomic regions extending 200 kb both upstream and downstream of significant SNP loci were systematically analyzed to identify genes associated with oil content in the soybean reference genome Williams 82 (http://www.soybase.org/). The selected window size is based on established linkage disequilibrium patterns in soybean, aiming to encompass both cis-regulatory elements and proximal genes that affect the trait of interest. A thorough functional annotation of the identified candidate genes was conducted by integrating various bioinformatic resources. These included The Arabidopsis Information Resource (TAIR), Gene Ontology (GO), Protein Families Database (PFAM), Protein Analysis Through Evolutionary Relationships (PANTHER) databases, and KOG (clusters of orthologous groups for eukaryotic entire genomes) annotation. The multi-database approach facilitated a comprehensive characterization of gene function, evolutionary conservation, protein domain structure, and metabolic pathway involvement, thereby aiding in the identification of biologically plausible candidates that regulate soybean seed oil accumulation.

3. Results

3.1. Statistical Analysis of Phenotypic Data

Field trials were conducted over two years (2021 and 2023) to evaluate the oil content of 341 soybean germplasm accessions. During the years 2021 and 2023, the natural soybean population exhibited notable phenotypic variation in seed oil content. The mean oil content increased from 18.43% in 2021 (SD = 2.12; CV = 11.48%) to 20.37% in 2023 (SD = 1.75; CV = 8.61%). The two-year average was documented at 19.35% (SD = 1.94; CV = 10.01%), as depicted in Figure 1A–C. Density plots demonstrated a unimodal and normal distribution across all years; however, the distribution in 2023 exhibited greater compactness compared to 2021 (Figure 1A–C). A one-way ANOVA revealed significant differences in oil content across the three treatment groups (2021, 2023, and Average; 341 accessions were evaluated each year), as demonstrated in the violin plot (Figure 1D). Tukey’s HSD test indicated the average performance across the two years. The analysis of oil content revealed that the distribution of oil content varied significantly across different years. A statistically significant positive correlation (r = 0.84, p < 0.001) was observed between oil content values recorded in 2021 and 2023 (Figure 1E), indicating a high level of consistency in genotype performance across the years. Ridge density plots (Figure 1F) illustrate the variations in oil content distributions across the years, clearly showing the rise in central tendency and the reduction in variability from 2021 to 2023.

3.2. Soybean Oil Content Across Northern China’s Growing Regions (2021–2023)

Variations in soybean oil content from 2021 to 2023 showed consistent and significant improvements across germplasms from all five main soybean-growing regions in Northern China (Figure 2A,B). The germplasm collection displayed strong regional differences, with most accessions coming from three northern provinces: Jilin (153 accessions, representing 44.87% of the total), Heilongjiang (126 accessions, 36.95%), and Liaoning (47 accessions, 13.78%). In contrast, Inner Mongolia and Xinjiang contributed only a small number of accessions (Figure 2A). This phenotypic regional distribution of data indicated specific patterns of oil content improvement across the evaluated areas.

The violin plots clearly show a steady upward trend in oil content distribution among germplasms from all regional origins, with 2023 data showing higher average values and more tightly clustered dispersion compared to the more spread-out patterns seen in 2021 (Figure 2B). Visualizations of both absolute and percentage changes reveal a consistent positive trend across all germplasm origins, indicating widespread improvement in oil content across the genetic diversity of soybean resources in Northern China (Figure 2D).

Germplasm from Heilongjiang showed a significant average increase of 2.03% in oil content, corresponding to an 11.65% relative gain (Figure 2B,C). This gain was present in 98.33% of the 126 assessed accessions. Materials from Inner Mongolia showed an average increase of 1.75%, which corresponds to a 9.38% relative gain. Furthermore, the enhanced phenotype was observed in all examined samples, with a 100% increase in oil content (Figure 2C). Accessions originating from Jilin demonstrated an average absolute increase of 1.73%, representing a 9.56% relative improvement. Significantly, 96.5% of the 153 accessions displayed the improved oil accumulation trait. Germplasm from Liaoning showed an average absolute increase of 2.08%, corresponding to a 12.58% relative gain, with this phenotype consistently expressed across all 47 evaluated accessions. Although their representation within the collection was limited, materials from Xinjiang exhibited a superior capacity for oil accumulation, with a mean absolute increase of 2.28%, resulting in an 18.78% relative gain. This phenotype was observed in 100% of the subset of accessions.

3.3. Genome-Wide Association Analysis Results

Genotyping was conducted using resequencing data from 341 soybean germplasm sources, resulting in 1,048,576 high-quality SNP sites across the 20 soybean chromosomes. The linkage disequilibrium (LD) decay plot is illustrated in Figure 3A. The R² value (y-axis) decreases to 50% of its maximum within 1.8 kb (x-axis), indicating substantial genetic diversity in the examined soybean population. The rapid LD decay suggests that the mapping resolution is high, facilitating more accurate identification of causal variants associated with specific traits. Population structure analysis was performed using fastSTRUCTURE with K values ranging from 1 to 10. The optimal number of subpopulations was determined using cross-validation error analysis (Figure S1), which revealed the lowest error at K = 2, indicating the presence of two major ancestral groups within the germplasm panel. However, the gradual increase in cross-validation error beyond K = 2 and the absence of sharp transitions suggest a complex population structure with continuous genetic variation rather than discrete subgroups. Furthermore, the three-dimensional principal component analysis (PCA) plot (Figure 3B) illustrates that the first three principal components (PC1, PC2, and PC3) span a range of −500 to 500. The red dots represent the 341 soybean accessions. Their continuous distribution without clear clusters confirms the absence of distinct subpopulation structure yet implies a continuous pattern of genetic variation potentially linked to geographical origins or ecological adaptations. The scree plot from the PCA analysis demonstrates the variance attributed to each principal component (Figure 3C). The first component (PC1) accounts for approximately 6% of the total genetic variance, while PC2 and PC3 account for progressively smaller proportions (approximately 4–5% and 3–4%, respectively). The gradual decline in explained variance, as opposed to sharp decreases, is characteristic of populations exhibiting complex genetic architecture, where variation is distributed across multiple dimensions rather than concentrated in a few major axes. Figure 3D displays a kinship matrix heatmap accompanied by hierarchical grouping. The primarily yellow hue, accented by red patches along the diagonal, signifies poor relatedness among the 341 soybean accessions, thus affirming the substantial genetic diversity within the mapping population. The results collectively demonstrate that the soybean population utilized in this work has significant genetic variety with limited population structure, rendering it appropriate for association mapping to uncover genomic areas associated with oil content.

3.4. Genome-Wide Association Analysis and Candidate Gene

Genomic DNA from 341 soybean varieties was sequenced using the Hi-Seq 2000 platform at an average depth of 10×. Genome-wide association analysis for oil content identified 119 significant SNP loci exceeding the threshold of −log₁₀(p-value) = 6.2. Q–Q plots confirmed the appropriateness of the MLM model across both years (2021 and 2023). Manhattan plots revealed a continuous linear distribution of significant SNPs, suggesting that the identified markers were reliable and indicating that soybean seed oil content is likely controlled by multiple minor-effect polygenes (Figure 4). In 2021, 51 significant loci were detected, with 48 clustered on chromosome 8 and the remainder on chromosomes 11 (2) and 20 (1). Chromosome 8 contained two distinct SNP groups with opposing effects: 27 loci (8,038,113–9,503,562 bp; average p-value: 8.85 × 10⁻⁸) associated with reduced oil content, and 21 loci (8,439,205–9,501,912 bp; average p-value: 4.78 × 10⁻⁸) associated with increased oil content. The 2023 data yielded 30 significant associations distributed across chromosomes 1 (7), 8 (20), and 13 (3). Analysis of two-year average phenotypic data identified 34 significant SNP loci across chromosomes 2, 4, 5, 7, 8, and 11, with chromosome 8 containing the majority (29) of these associations (

Table 1. Significant SNP markers related to oil content.

Year	Chr	Physical Range of Significant SNP Loci		p Value	Effect	Number of Significant SNP Loci	Note
		Start	End
2021	8	8,038,113	9,503,562	8.85 × 10−8	−1.38	27	Reduce oil content
	8	8,439,205	9,501,912	4.78 × 10−8	1.45	21	Increase oil content
	11	11,087,536	11,087,536	1.79 × 10−7	0.89	1	Increase oil content
	11	11,103,604	11,203,604	0.02	−0.86	1	Reduce oil content
	20	15,764,307	15,764,307	2.80 × 10−7	−1.01	1	Reduce oil content
2023	1	10,828,034	10,841,167	1.91 × 10−7	−1.08	3	Reduce oil content
	1	10,537,841	10,904,178	2.46 × 10−7	1.07	4	Increase oil content
	8	9,005,247	9,079,037	1.41 × 10−9	−1.58	7	Reduce oil content
	8	9,005,430	9,078,617	2.94 × 10−8	1.45	12	Increase oil content
	8	42,038,411	42,038,411	2.19 × 10−7	1.10	1	Increase oil content
	13	45,137,495	45,137,495	1.86 × 10−7	−1.12	1	Reduce oil content
	13	45,140,366	45,163,611	1.56 × 10−7	1.13	2	Increase oil content
Average	2	14,714,159	14,714,159	1.16 × 10−7	−0.97	1	Reduce oil content
	4	15,518,980	15,518,980	2.16 × 10−7	−1.08	1	Increase oil content
	5	16,157,283	16,157,283	1.84 × 10−7	3.35	1	Reduce oil content
	7	16,529,749	16,529,749	1.12 × 10−7	0.94	1	Increase oil content
	8	8,432,800	9,501,864	6.65 × 10−8	−1.32	13	Reduce oil content
	8	7,801,179	9,501,912	2.63 × 10−8	1.37	13	Increase oil content
	8	42,039,721	42,039,740	5.65 × 10−8	1.13	2	Increase oil content
	8	42,039,735	42,039,735	1.80 × 10−7	−1.03	1	Decrease oil content
	11	33,092,710	33,092,710	2.42 × 10−7	1.04	1	Increase oil content

).

The presence of SNPs with opposing effects within overlapping genomic regions on chromosome 8 (8,038,113–9,503,562 bp and 8,439,205–9,501,912 bp) suggests several possible explanations that warrant further investigation. These opposing effects result from distinct haplotype blocks within this region, where different combinations of linked alleles create contrasting phenotypic outcomes depending on the specific allelic configuration inherited by individual accessions. The opposing effects indicate epistatic interactions between genes within this chromosomal region, where the phenotypic effect of one locus is modified by allelic variants at nearby loci. Such epistatic relationships are common in quantitative traits and could explain how closely linked SNPs produce contrasting effects on oil accumulation. The overlapping physical positions of these SNP groups (with a 400 kb overlap) suggest that they regulate a domain or involve functionally related genes that interact to fine-tune oil biosynthesis pathways. The opposing effects reflect the presence of multiple functional variants that affect different aspects of lipid metabolism within the same genomic region. For instance, some variants influence fatty acid biosynthesis, while others affect oil storage or mobilization, resulting in either a net-positive or -negative effect on total seed oil content, depending on the specific combination of alleles present. The consistent identification of significant loci on chromosome 8 across multiple environments highlights this region as particularly important for soybean oil content regulation and provides promising targets for marker-assisted selection in breeding programs aimed at modifying seed oil profiles. However, the complex pattern of opposing effects within this region suggests that breeding strategies should consider haplotype-based selection rather than individual SNP markers to capture the full genetic architecture and avoid unintended consequences from epistatic interactions.

3.5. Candidate Gene Screening for Association Analysis

Soybean oil content was found to be substantially associated with 29 SNP locations on chromosome 8. Specifically, in 2021, 19 SNP sites were identified as being strongly correlated with soybean oil content, and in 2023, this number decreased to 2 (

Table 2. Information on SNP loci associated with soybean seed oil content on chromosome 8.

Year	Gene	Position (bp)	p Value	MAF	DNA Sequence Variation	Protein Sequence Variation	PROVEAN Score
2021	Glyma.08G110000	8,439,205	2.51 × 10−7	0.08	A739C	T247P	0.031
	Glyma.08G117400	9,074,920	8.62 × 10−11	0.05	G682T	G228C	1.450
	Glyma.08G117600	9,501,436	6.76 × 10−12	0.05	G698A	A233V	1.233
	Glyma.08G123500	9,501,620	2.36 × 10−7	0.06	G2281A	V761I	−0.200
	Glyma.08G123500	9,501,695	5.54 × 10−8	0.06	G1650A	E550D	−0.667
	Glyma.08G123500	9,074,920	2.61 × 10−7	0.06	G1567C	E523Q	−0.200
	Glyma.08G123500	9,074,920	2.19 × 10−7	0.06	T1555C	C519R	−0.017
	Glyma.08G123500	9,501,695	6.45 × 10−9	0.06	A1508G	E503G	−1.633
	Glyma.08G123500	9,501,912	2.03 × 10−8	0.06	G1490C	L497P	−0.083
	Glyma.08G123500	9,501,456	3.40 × 10−8	0.06	T1488G	F496L	0.933
	Glyma.08G123500	9,501,556	1.11 × 10−7	0.06	T1388A	I463N	0.592
	Glyma.08G123500	9,501,564	6.32 × 10−8	0.06	T1380G	F460E	−1.092
	Glyma.08G123500	9,501,586	2.39 × 10−7	0.06	T1358A	V453E	0.200
	Glyma.08G123500	9,501,695	9.17 × 10−10	0.06	A1249G	K417E	0.058
	Glyma.08G123500	9,501,776	3.87 × 10−8	0.06	G1168A	D390N	0.133
	Glyma.08G123500	9,501,864	9.23 × 10−8	0.06	C1080A	D360E	−0.033
	Glyma.08G123500	9,501,912	4.54 × 10−8	0.06	C1032G	N344K	0.350
	Glyma.08G123500	9,502,316	2.47 × 10−7	0.05	T628G	S210A	0.383
2023	Glyma.08G117400	9,054,741	4.94 × 10−12	0.05	G682T	G228C	1.450
	Glyma.08G117600	9,074,920	8.71 × 10−14	0.05	C698T	A233V	1.233
Average	Glyma08G117400	9,054,741	7.46 × 10−11	0.05	G682T	G228C	1.450
	Glyma08G117600	9,074,920	5.44 × 10−13	0.05	C698T	A233V	1.233
	Glyma08G123500	9,500,663	2.60 × 10−7	0.06	G2281A	V761I	−0.200
	Glyma08G123500	9,501,454	1.47 × 10−7	0.07	T1490C	L497P	−0.083
	Glyma08G123500	9,501,456	7.61 × 10−8	0.06	T1488G	F496L	0.933
	Glyma08G123500	9,501,695	3.88 × 10−9	0.06	A1249G	K417E	0.058
	Glyma08G123500	9,501,864	1.66 × 10−7	0.07	C1080A	D360E	−0.033
	Glyma08G123500	9,501,912	1.84 × 10−7	0.07	C1032G	N344K	0.350

). By averaging the data from 2021 and 2023, the GWAS analysis identified eight SNP locations that showed a strong association with soybean oil content. Multiple candidate genes involved in regulating oil content were identified by further examination of the SNP variations (Table 2).

The gene with the most significant associations in both the annual and combined analyses was Glyma.08G123500. SNPs appear to have altered the protein’s function in this gene. With p-values ranging from 3.88 × 10⁻⁹ to 1.84 × 10⁻⁷, significant SNPs potentially producing alterations, such as L497P, F496L, K417E, D360E, and N344K, were repeatedly identified. A G228C amino acid change occurred at position 9,054,741 bp due to a substantial SNP in the gene Glyma.08G117400 (Table 2). In both 2023 and combined analyses, this variant was highly significant (p = 4.94 × 10⁻¹² and p = 7.46 × 10⁻¹¹, respectively). The most statistically significant association across all discovered associations in 2023 and the combined analysis was an SNP at position 9,074,920 bp resulting in a predicted A233V substitution (Glyma.08G117600, p = 8.71 × 10⁻¹⁴, p = 5.44 × 10⁻¹³). An SNP at 8,439,205 base pairs (bp), which potentially alters the amino acid sequence to T247P, was found to be strongly associated with oil concentration in 2021 (p = 2.51 × 10⁻⁷).

PROVEAN functional impact scores ranged from −1.633 to 1.450, with potentially deleterious variants including E503G (−1.633) and F460E (−1.092) in Glyma.08G123500, while variants in Glyma.08G117400 (G228C) and Glyma.08G117600 (A233V) showed positive scores (1.450 and 1.233, respectively) but maintained strong temporal consistency between datasets. These notable SNPs have minor allele frequencies (MAFs) between 0.05 and 0.08, indicating that they are relatively rare variants in the population. The consistent detection of these SNPs across various contexts, especially in Glyma.08G123500, Glyma.08G117400, and Glyma.08G117600, suggests that they have stable effects on oil content and could potentially be used as markers in breeding programs to aid selection.

Based on the comprehensive genetic variation analysis of four soybean genes on chromosome 8, this study reveals diverse functional roles and varying impacts of genetic variants on protein function. Glyma.08G110000, encoding a hydroxycinnamoyl coenzyme A-quinate transferase involved in secondary metabolism (GO:0016747), shows high expression in reproductive tissues and contains a single neutral variant (T247P, PROVEAN score: 0.031). Glyma.08G117400, a PPR repeat protein involved in RNA metabolism and organellar function (GO:0005515), demonstrates consistent variants across years with neutral predicted effects (PROVEAN score: 1.450) and broad tissue expression. Glyma.08G117600, a WD40 repeat scaffolding protein with protein binding activity (GO:0005515), exhibits the most significant association (p = 5.44 × 10⁻¹³) and high expression in seeds and pods, despite carrying apparently neutral variants (A233V, PROVEAN score: 1.233). Most notably, Glyma.08G123500, a receptor-like kinase with protein kinase activity (GO:0004672) and signal transduction functions, contains 15+ variants with PROVEAN scores ranging from deleterious (−1.633) to neutral (0.933), and is most highly expressed in metabolically active root and nodule tissues (7.45 and 5.34 FPKM, respectively). This complex genetic architecture suggests Glyma.08G123500 as the most promising candidate for functional validation studies.

4. Discussion

4.1. Phenotypic Variation and Environmental Effects

The significant increase in mean oil content from 18.43% in 2021 to 20.37% in 2023, along with a reduced coefficient of variation (11.48% to 8.61%), suggests that environmental conditions in 2023 were more favorable for oil accumulation across the soybean germplasm collection. This improvement was remarkably consistent across all geographical regions, with enhancement phenotypes observed in 96.5–100% of accessions depending on origin. The strong positive correlation (r = 0.84, p < 0.001) between years demonstrates the stability of genetic effects despite environmental variation, supporting the reliability of our phenotypic data for association mapping. The regional analysis revealed interesting patterns, with Xinjiang germplasm showing the highest relative improvement (18.78%) despite limited representation in the collection. The consistent improvement across Northeastern China’s major soybean-producing provinces (Jilin, Heilongjiang, and Liaoning) indicates that the favorable environmental conditions in 2023 were widespread across the region [51,52].

4.2. Population Structure and Genetic Diversity

A natural population of 341 soybean varieties was utilized to assess the oil content of soybean seeds in 2021 and 2023, and a GWAS analysis was conducted using 1,048,576 SNP markers acquired through resequencing. The identification of 119 significant SNP loci across multiple chromosomes confirms that soybean seed oil content is controlled by a polygenic architecture, consistent with earlier quantitative genetic studies. The concentration of significant associations on chromosome 8 (especially in regions 8.0–9.5 Mb and around 42.0 Mb) across various environments strongly indicates the presence of major QTLs in these areas. The consistency of these associations over the years (2021, 2023, and combined analysis) offers strong evidence for the stability of genetic effects on oil content. Notably, the detection of SNP clusters with opposing effects within the same region on chromosome 8 suggests either multiple linked genes with different functions or allelic variants of the same genes affecting oil metabolism. This complexity is common in metabolic traits and likely reflects the intricate regulatory networks that control lipid biosynthesis and accumulation in seeds [53,54].

4.3. Candidate Gene Analysis and Functional Implications

The identification of Glyma.08G123500, Glyma.08G117400, and Glyma.08G117600 as consistently associated genes across multiple analyses provides strong evidence for their potential role in regulating oil content. Functional domain analysis reveals that Glyma.08G123500 (also known as Glyma.08g13040) encodes a multi-domain protein featuring an NB-ARC domain (a nucleotide-binding adaptor shared by APAF-1, R proteins, and CED-4), a protein tyrosine and serine/threonine kinase domain, and leucine-rich repeats (LRRs). These characteristics are typical of NBS-LRR resistance proteins, which primarily function in plant immune responses and stress signaling pathways [55,56]. The detection of multiple amino-acid-changing variants within Glyma.08G123500, including potentially deleterious mutations (E503G with a PROVEAN score of −1.633 and F460E with a score of −1.092), suggests that variants affecting the NB-ARC nucleotide-binding activity or kinase function could substantially impact protein signaling capacity. Although NBS-LRR proteins are primarily recognized for their role in pathogen resistance, emerging evidence indicates that some family members may also regulate metabolic processes and stress responses, potentially influencing seed development and oil accumulation through coordinated stress signaling with lipid biosynthesis during seed filling [57]. This finding aligns with recent studies that demonstrate the multifunctional nature of NBS-LRR proteins, extending beyond disease resistance. For instance, Wu et al. [58] reported that certain NBS-LRR genes in Arabidopsis are involved in regulating developmental processes. Specifically in soybeans, previous research has suggested that stress response pathways and lipid metabolism may be interconnected during seed development [31]. The amino acid substitutions identified (L497P, F496L, K417E, D360E, and N344K in Glyma.08G123500; G228C in Glyma.08G117400; A233V in Glyma.08G117600) represent non-conservative changes that could markedly alter protein structure and function. Structural analysis suggests that these substitutions may affect critical functional domains, potentially altering protein–protein interactions or enzymatic activity [26]. The consistent detection of these variants across different environmental conditions and years reinforces their biological significance, suggesting that they represent stable genetic factors rather than spurious associations [11]. The relatively low minor allele frequencies (0.05–0.08) of these variants indicate that they represent rare but functionally significant alleles within the germplasm panel. This pattern is consistent with previous observations that variants with large phenotypic effects often occur at low frequencies in natural populations due to balancing selection or recent evolutionary origin [59]. Such rare variants with substantial effects could be particularly valuable for breeding programs seeking to modify oil content, as they may provide novel allelic diversity not commonly available in elite breeding lines [43]. The functional implications of these findings extend beyond the effects of individual genes. The clustering of significant associations within a relatively narrow chromosomal region (approximately 1.5 Mb on chromosome 8) suggests the presence of a complex genetic architecture involving multiple linked genes or regulatory elements that collectively influence oil content [20]. This pattern may reflect the presence of ancient chromosomal inversions or other structural variants that maintain beneficial allelic combinations in linkage disequilibrium [60].

4.4. Quantitative Comparison with Prior GWASs

Comparative analysis with previous soybean oil content GWASs [61,62] revealed that 47 of our 119 significant SNPs (39.5%) were located within previously reported QTL regions, including overlaps with Seed oil 11-g2 [63] and other major oil content loci [49,50], while 72 SNPs (60.5%) represent novel associations not captured in prior studies. Specifically, within the critical Glyma.08G123500 locus, 15 of the 18 identified SNPs (83.3%) represent previously uncharacterized variants, with only 3 showing positional overlap with variants reported in previous association studies [62], though none of the overlapping variants showed identical allelic effects or population frequencies.

Our Northeast China population (n = 341) exhibited a distinct genetic structure compared to previous studies [62,64], with an average genetic diversity (π) of 2.1 × 10⁻³ and population differentiation (FST) values of 0.23–0.31 relative to major U.S. and Brazilian soybean collections used in prior GWAS analyses [65]. Environmental specificity analysis demonstrated that 31 of the 119 significant associations (26.1%) showed consistent detection across both 2021 and 2023 growing seasons under Northeast China climatic conditions.

The integration of PROVEAN functional impact analysis identified eight potentially deleterious variants (PROVEAN score < −1.0) within candidate genes that were not functionally characterized in previous studies, providing novel insights into the molecular mechanisms underlying oil content variation. Comparative marker density analysis showed that our study achieved 3.1-fold higher resolution (1 SNP per 456 bp) compared to previous array-based GWASs (1 SNP per 1400–2100 bp) [61,62,63], enabling detection of fine-scale associations and improved candidate gene localization.

4.5. Breeding and Selection Implications

The identification of stable genetic markers associated with oil content provides valuable tools for marker-assisted selection in soybean breeding programs. The consistent detection of chromosome 8 associations across environments makes these regions particularly attractive targets for breeding efforts aimed at modifying seed oil profiles. The relatively small effect sizes of individual SNPs (ranging from −1.58 to +3.35 percentage points) suggest that pyramiding multiple favorable alleles will be necessary to achieve substantial improvements in oil content. The regional variation in oil content enhancement observed in our study also suggests that different germplasm sources contribute unique favorable alleles. The superior performance of Xinjiang germplasm, despite its limited representation, suggests that expanding genetic diversity through the inclusion of germplasm from diverse geographical origins could provide additional genetic variation for improving oil content.

4.6. Study Limitations and Future Directions

In contrast to classical linkage analysis, GWASs can utilize existing natural populations as subjects, thereby conserving time and resources [59]. Nonetheless, it possesses several inherent limitations that must be considered when interpreting association results. Population structure remains a primary concern, as cryptic relatedness and subpopulation stratification can lead to spurious associations and inflated Type I error rates [66,67]. Linkage disequilibrium patterns, influenced by demographic history, selection pressure, recombination rates, and population bottlenecks, can confound the precise localization of causal variants, particularly in regions of extended LD where multiple linked variants show similar association signals [68]. Additionally, the “missing heritability” problem persists, where identified SNPs explain only a fraction of the phenotypic variance, potentially due to rare variants with large effects, structural variations not captured by SNP arrays, or epistatic interactions between loci that are not adequately modeled in single-marker association tests [69]. The resolution for candidate gene identification is further limited by the density of markers relative to local LD structure, and the assumption that the causal variant is in linkage disequilibrium with the tested marker does not hold in all genomic regions [70]. To address these methodological challenges, advanced statistical approaches such as the Fixed and Random Model Circulating Probability Unification (FarmCPU) algorithm have been developed, which iteratively optimizes both fixed and random effects to better control population structure and kinship while maintaining statistical power for detecting true associations [71]. FarmCPU and similar multi-locus models can reduce false positive rates by accounting for the confounding effects of population stratification more effectively than traditional single-locus approaches, though they require careful parameter tuning and validation across different populations and traits [71,72]. To achieve more precise study outcomes, we can minimize false positives by augmenting the population size. We employ a mixed linear model (MLM) that simultaneously incorporates individual kinship (K) and group structure (Q) as covariates (Q + K), therefore successfully diminishing the false positive rate [73]. GWASs utilizing mixed linear models are now extensively applied in both botanical and zoological systems [74]. GWASs are increasingly employed in agricultural research; for instance, the genetic foundations of many seed-related features have been identified in soybean, with multiple candidate genes predicted using GWASs [15,60,75]. A total of 92 QTLs influencing quality variables were identified across three environments, with 14 QTLs associated with crude oil content [76]. Additionally, 32 candidate genes associated with soybean quality were identified. To date, 393 QTL loci associated with soybean seed fat content traits have been cataloged on the SoyBase website (http://www.soybase.org), spanning 20 chromosomes [77]. This collection comprises 298 QTLs derived from the RIL population and 95 QTLs from the GWAS population [78].

Future research should focus on experimental validation of Glyma.08G123500 through CRISPR/Cas9-mediated functional studies and biochemical characterization of the identified protein variants, particularly the deleterious E503G and F460E substitutions, to confirm their causal role in oil content variation. The unexpected identification of an NBS-LRR resistance protein as a major regulator of seed oil content underscores the intricate relationships between plant immune responses and metabolic processes, suggesting that stress signaling pathways coordinate with lipid biosynthesis during seed development. These findings provide valuable molecular markers for marker-assisted selection in soybean breeding programs, opening new avenues for understanding the genetic architecture underlying economically important seed quality traits. This has broader implications for crop improvement strategies that leverage the pleiotropic effects of stress response genes on metabolic phenotypes.

5. Conclusions

Our analysis identified 119 significant SNP loci associated with oil content, with chromosome 8 emerging as a key region containing 77 significant associations across multiple environments. The candidate gene Glyma.08G123500 showed strong and consistent associations, with several SNPs causing amino acid substitutions that may alter protein function and affect oil biosynthesis. Additional candidates, Glyma.08G117400 and Glyma.08G117600, also warrant further investigation. The consistent overlap between our findings and previously reported QTLs confirms our mapping approach and emphasizes the role of chromosome 8 in regulating soybean oil content. Notably, we identified SNP clusters with opposing effects on oil content, indicating the complex genetic regulation of this trait. The significant increase in oil content across all Northern Chinese growing regions from 2021 to 2023 demonstrates the potential for continued improvement through targeted breeding. The SNP markers identified in this study offer valuable resources for marker-assisted selection to expedite the development of high-oil soybean varieties.