Plant material and growth conditions

A RIL population (‘Forrest’ × ‘Williams 82’) was developed for genetic mapping [24]. ‘Forrest’ (PI 548402) [25] was created from a cross between ‘Dyer’ and ‘Bragg’, developed by USDA [25], and ‘Williams 82’ (PI 518671) [26] from a cross of ‘Williams’ and ‘Kingwa’ [26] (Fig 1). The genetic map was based on 306 RILs and 2075 SNP markers [27,28]. Our previous published research mapped QTL for soybean seed protein, oil, isoflavone, and amino acids, and showed that the parents (Williams 82 and Forrest) were contrasting in seed protein, oil, isoflavones, and some amino acids [28–30].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Pedigree of the ‘Forrest’ × ‘Williams 82’ recombinant inbred line (RIL) soybean population used in this study.

https://doi.org/10.1371/journal.pone.0331214.g001

‘Forrest’ was originally developed for resistance to soybean cyst nematode (Heterodera glycines) and has been widely utilized in genetic mapping studies due to its agronomic resilience and seed composition traits. ‘Williams 82’ (PI 518671) [26], a cultivar released for resistance to Phytophthora sojae, serves as the reference genome in soybean genomics. ‘Forrest’ was released in 1973 by the USDA-ARS and Mississippi Agricultural and Forestry Experiment Station in cooperation with the Tennessee Agricultural Experiment Station [25]. It is a mid-maturity group V cultivar with a determinate growth habit, purple flower color, and yellow seeds. It has shown good performance under disease pressure and southern U.S. growing conditions. Forrest has been widely used in QTL studies due to its resistance traits and unique seed composition characteristics. ‘Williams 82’ (PI 518671), released in 1988 by the USDA-ARS and Illinois Agricultural Experiment Station [26], is a maturity group III cultivar with a determinate growth habit, purple flowers, and yellow seeds. It is used as the reference genome in soybean genomics due to its genetic stability and broad adaptability. Our previous published research mapped QTL for soybean seed protein, oil, isoflavone, amino acids, and showed that the parents (Williams 82 and Forrest) were contrasting in seed protein, oil, isoflavones, and some amino acids (included threonine, serine, proline, glycine, alanine, cysteine, valine, methionine, phenylalanine, lysine, tryptophan), as reported elsewhere [28–30].

While reports specifically comparing seed composition between these two cultivars are limited, our previous work has identified significant genetic differences in traits such as seed isoflavones [27], protein, oil, sugars, and tocopherols [29], and mineral nutrients [28,30]. These findings support the use of this RIL population for studying seed nutritional traits. The experiments were conducted on university research farms as described in Knizia et al. (2021) [27]. Two field experiments were conducted using the ‘Forrest’ × ‘Williams 82’ RIL population. The first experiment was carried out in 2018 at a Farm in Spring Lake, NC (35.17° N, 78.97° W) rented by Fayetteville State University. The second experiment was conducted in 2020 at the Southern Illinois University Agronomy Research Center in Carbondale, IL (37° N, 89° W). Both sites are managed by the university research farms with standard agronomic practices and irrigation capabilities. Experimental protocols, field layout, and management practices followed those detailed in Knizia et al. (2021) [27].

Seeds were planted on a 75 cm row-spacing, and the growth conditions and field management were conducted as previously reported by others [27,28]. In the current experiment, RILs showed a wide range of N, S, and P, and some RILs showed even higher concentrations of N, S, and P than the two parents. This means that RIL population showed Transgressive Segregation where the progeny shows more genetic variation and variation in gene expression than their parents. Also, this indicated that N, S, and P content trait in soybean is a complex polygenic trait, as previously reported [29], where QTL mapping for different seed nutrients, including N, S, and P using various mapping biparental populations. Therefore, QTL associated with N, S, and P were pursued.

Analysis for seed N and S

The concentrations of N and S in the matured seed were determined by using ground and dried seeds using a Laboratory Mill 3600 (Perten, Springfield, IL USA). Nitrogen and S concentrations measurements in mature seeds were conducted using a 0.25 g ground-dried sample, and were combusted in an oxygen atmosphere at 1350 ºC, converting elemental N and S into N₂ and SO₂, respectively. Gases of N₂ and SO₂ were then passed through infrared cells and N and S were measured by an elemental analyzer using thermal conductivity cells (LECOCNS-2000 elemental analyzer, LECO Corporation, St. Joseph, MI USA) as previously detailed [31,32].

Phosphorus measurement

Mature seeds at R8 were collected and analyzed for P concentrations using spectrophotometer. The concentration of P was determined by the yellow phosphor-vanado-molybdate complex as detailed elsewhere [33]. In short, a sample of 2 g of a dried ground seed was ashed, and then, 10 ml of 6 M HCl was added and placed in a water bath to bring the solution to dryness. Then, 2 ml of 36% v/v HCl were added, and the sample was boiled. The samples were boiled again by adding 10 ml of distilled water. The sample, then was transferred to a 50-mL volumetric flask, diluted to 50 mL with distilled water, filtered, and a volume of 2 ml of filtrate was first discarded and the rest was kept for P analysis. A volume of 5 ml of 5 M HCl and 5 ml of ammonium molybdate–ammonium metavanadate reagent were added to the filtrate, and the mixture was diluted with distilled water to 50 ml. Ammonium molybdate–ammonium metavanadate was prepared by dissolving 25 g of ammonium molybdate and 1.25 g of ammonium metavanadate in 500 ml of distilled water. Dihydrogen orthophosphates was used to prepare a P standard curve (0–50 μg/ml of P). A Beckman Coulter DU 800 spectrophotometer was used to measure the P concentrations at 400 nm.

DNA isolation, SNP genotyping, and genetic map construction

Genomic DNA of the RIL population and their parents was extracted according to others and our previous published research [27,28,30,34,35]. The RIL population was genotyped with BARCSoySNP6K Illumina Infnium BeadChips [35, https://www.soybase.org/tools/snp50k/]. Genotyping was conducted in the Soybean Genomics and Improvement Laboratory, USDA-ARS, Beltsville, MD, USA. A threshold of 2.5 for LOD, and a maximum genetic distance of 50 cM to group markers were used. The linkage groups were assigned to corresponding soybean chromosomes as described in Soy-Base [36,37] (https://www.soybase.org/about/lgs_and_chromosomes/).

N, S, and P QTL detection, candidate genes, and statistics

Detection of QTL and statistical analysis were performed as previously described [27,28]; the broad sense heritability (H²) analysis from two-way ANOVA was performed using the equation:

(1)

As described in details by others and by our previous published work [27,28,30,38]. The significance level was conducted using R package car (type II Wald chi-square tests) (R Software, accessed on June 15, 2023) [39]. To identify QTL for seed N, S, and P concentrations in RIL population, we used Composite Interval Mapping (CIM) methods of Win-QTL Cartographer 2.5 (Windows QTL Cartographer 2.5. Department of Statistics, North Carolina State University, Raleigh, NC; http://statgen.ncsu.edu/qtlcart/WQTLCart.htm) [40]. The default parameters of WinQTL Cartographer were selected (Model 6, 1 cM step size, 10 cM window size, 5 control markers, and 1,000 permutations threshold) [40]. Chromosomes were drawn using MapChart 2.2 [41]. Designation of QTL in the two years were done following the approach used by others [28,41]. We used CIM method because it is a widely accepted and robust method for detecting QTL with moderate to high power and resolution. We acknowledge that alternative methods such as Multiple QTL Mapping (MQM)or Inclusive Composite Interval Mapping (ICIM) may offer complementary insights, particularly in separating closely linked QTL or modeling epistatic interactions. However, the current study focused exclusively on CIM, and we did not perform MQM analyses. The candidate genes, within identified QTL for N, P, and S content in soybean seeds, were annotated using SoyBase Genome Browser (glyma.Wm82.gnm4).

A randomized complete block design (RCBD) was used for these experiments, with three replicates per genotype. This experimental layout was adapted from Knizia et al. [27], and followed standard field protocols for soybean QTL mapping studies. Since Hurricane Florence caused damage to the experimental site in Spring Lake, NC, in 2018, phenotypic and QTL analysis for that location was limited to 187 undamaged RILs (n = 187). However, the 2020 experiment in Carbondale, IL, was not affected, and data were collected from the full population of 306 RILs (n = 306). Analysis of Means procedure was conducted using Proc Means in SAS. Correlations were conducted by SAS using PROC REG (SAS, Statistical Analysis Systems, Cary, NC, USA, 2002–2012) [42].

QTL analysis for N, P, and S

The significant effect of line and location is due to the genotypic differences among lines and the contribution of the environment to the variation of the trait. The high heritability of 91.7% of seed N trait variation is due to genetics, followed by 48.2% for S seed concentration trait, and finally an inheritance of close to zero for seed P concentration trait. The very low inheritance, recorded in seed P may indicate the strong effect of environmental factors on the trait and very little effects due to genetic differences. Also, the low heritability that was shown in seed P could be due to complex quantitative trait, gene-to-gene interactions (epistasis effect), and significant gene by environment interactions with P, and suggest limitations of these QTL use in the breeding selection [28]. It was reported on other minerals that the non-genetic factors could be extremely high [28,29]. They further added that the low heritability for P could be due to the interaction between the trait and environment. Also, the large interaction of gene with environment for P QTL may require a large number of RILs across locations and across years to obtain significant QTL with high inheritance [28,30]. This will require further research before final conclusions are made. The positive correlations between N, S, and P nutrients in both years and at both locations indicated the symport relationship between these nutrient transport system. The negative correlation between N and P in 2018 indicated antiport relationship that can occur due to environmental factors, especially drought, heat, and soil conditions. Positive and negative mineral relations were previously observed [11,12,43–45].

To our knowledge previous genomic research, conducted on minerals, was mostly conducted on leaves, roots, or shoot [3,15,16], and not on seed N, S, and P nutrients accumulation in soybean seeds [3,17,18]. For example, searching SoyBase (https://www.soybase.org/search/index.php?searchterm=Nitrogen±and±Phosphorus±and±Sulfur&list=bi_parental_qtl_listview) revealed that only QTL and molecular markers related to N, P, and S concentrations in shoot tissue were identified [17]. They were able to identify soybean shoot QTL (one QTL for P (qPHO001), one QTL for N (qNIT001), and one QTL for S (qSUL001). They explained that QTL clustering of P, K, Mg, C, N, and S indicated physiological and genetic relationships, and possible similar metabolic processes between these nutrients. Other researchers identified QTL related to leaf minerals and were able to identify, using 200 RILs, 6,366 SNPs markers that covered the whole genome, and 19 QTL and 3 candidate genes associated to leaf-related traits [19]. Previous research, using 92 F_5:7 (F5-derived) soybean RIL, a cross between MD 96–5722 (MD) and Spencer, and 5,376 SNP markers, were also able to detect QTL related to seed N, P, and S [23]. They identified N QTL qNIT001 on Chr 16 [23]. They were also identified P QTL qPHO001 on Chr16. The S QTL q SUL001 on Chr16 was also detected [23]. Genomic regions associated with seed N accumulation in soybean at R5, R6, and R7 growth stages were also identified [19]. They used a population of 101 F_6:8 (F6-derived) RILs, a cross between N87-984–16 × TN93–99 [19]. They detected QTL on Chr 2, 7, 8, 14, 15, 18, and the contribution to the phenotypic variation ranged from 5 to 11.6%.

It must be noted in our research that LOD for N QTL in 2020 was large. This is can be due to the following possible reasons: one is a larger population size (n = 306); unlike the NC site in 2018, which had reduced data due to hurricane damage (n = 187), the IL 2020 trial included the full RIL population. The increased number of data points likely enhanced statistical power, reducing residual variance and inflating LOD values; two, a greater trait variation and data quality, i.e., environmental conditions in Carbondale were more stable, and no external damage affected plant performance. As a result, trait expression was clearer, and stronger marker-trait associations could be detected, especially for seed N, which showed high heritability (H² = 0.917); third, a narrow confidence intervals and high additive effects: In several cases (e.g., qN-01-[IL-2020] and qN-03-[IL-2020]), the trait exhibited strong additive effects combined with tightly linked markers, further contributing to elevated LOD values.

For P, other researchers worked on genomic regions for other minerals such as P. For example, using 184 RILs, a cross between Kefeng No. 1 and Nanong 1138−2 soybean varieties, others [16] were able to identify QTL for P deficiency tolerance in leaves, roots, and shoots, and identified seven QTL associated with weight of fresh shoot, P contents in leaf and in root. Other also identified P QTL associated with shoot P accumulation [20–22]. The low P content heritability has four important aspects: the first aspect is the biological and agricultural relevance, i.e., seed P content remains a biologically and nutritionally important trait in soybean, especially due to its connection with seed vigor, germination, and nutritional quality. Even when heritability is low, understanding the genetic basis of this trait is critical for identifying genotypes or genomic regions responsive to environmental P variability; the second aspect is the precedent in literature, i.e., the low heritability for P-related traits in soybean has been reported previously, especially under single-environment studies or limited replication. For instance, Li et al. (2005) [16] and Zhang et al. (2016) [20] both reported variable P heritability depending on tissue type, environment, and developmental stage. These studies still identified meaningful QTL by using multi-location or repeated trials; similar to our approach; the third aspect is the detection of stable QTL, i.e., despite low overall heritability, we were able to identify several statistically significant QTL for seed P content in both environments (Table 3), with LOD values ranging from 2.5 to 9.4 and phenotypic variance explained (R²) up to 12.17%. This suggests that specific loci still contribute consistently across environments, even if the overall trait expression is environmentally labile; the fourth aspect is the justification through G × E, i.e., the low heritability itself is an important genetic insight. It indicates a potential for strong environmental modulation or G × E interactions, which could be exploited through genotype-specific agronomic strategies or P management practices. Moreover, inclusion of this trait helps distinguish which QTL are environment-stable versus those that are more plastic. Our research showed that, except for QTL detected on Chr 16 [23], 11 QTL reported here were not previously identified, therefore, they are novel.

Candidate genes and gene annotation for N P S

Candidate genes were identified for N, P, and S concentrations in soybean seeds were annotated using SoyBase Genome Browser (glyma.Wm82.gnm4) (Table 4 and Table 5). These annotated genes were located either within the identified QTL interval or vicinity areas. For N, more than 11 genes, including ammonium transporter, nodulin like proteins, glutamine-dependent NAD (+) synthetase, ammonium transmembrane transporters were identified in the functional loci, suggesting the functional association with N metabolism and accumulation and their genetic variants. These candidate genes may play roles in ammonium transport across cellular membranes, reversible transfer of an amino group between alanine and α-ketoglutarate and forming glutamate and root nodules of leguminous plants. The ammonium transporters (AMTs) are important genes involved in ammonium absorption and utilization in soybeans. The QTL underlying N concentration were identified on chromosome 2 and 10 in this research and explained more than 8% and 42% phenotypic variation, respectively. The candidate genes associated with these two loci, Glyma.02G043700 and Glyma.10G030800 were identified in the intervals. The same candidate genes GmAMT4.4 (Glyma.02G043700) and GmAMT4.6 (Glyma.10G030800) were also annotated, and genes expression of GmAMT family genes in the soybean plants, using a GWAS, was conducted by others [46]. Both candidate genes were among 16 GmAM paralog genes annotated in the study and GmAMT4.6 was confirmed to be related with the circadian rhythms in the transcription analysis [46]. Moreover, there were several candidate genes existing within the loci for the traits of P and S, suggesting the functional significance of these intervals for the nutrients in soybean. The candidate gene inositol-tetrakisphosphate 1-kinase 2-like isoform X1 (Glyma.04G088000) was annotated in Soybase involved in the metabolism of inositol phosphates as the signaling molecules. Protein phosphatase 2C family gene (Glyma.03G043675) appeared to be involved in the regulation of crucial steps in the cells by catalyzing the removal of phosphate groups from proteins. The candidate gene purple acid phosphatase 17 (Glyma.20G011900) under the QTL (qP-03) on chromosome 20 is involved in phosphate metabolism. The QTL (qP-03) explained more than 5% phenotypic variation. Zhu et al. (2020) [47] studied the dynamic changes of intracellular (leaf and root) and root-associated APase (purple acid phosphatase) activity under both Pi-sufficient and Pi-deficient conditions for a total of 38 purple acid phosphatase (GmPAP) members identified in the soybean genome. The relative expression levels of GmPAP17c (Glyma.20G011900) and other GmPAPs in the leaves and roots were analyzed at two P levels using qRT-PCR analysis. However, the transcript abundance of the GmPAP17c was detected at both P levels in the leaves, but not in the roots [47]. There were several candidate genes identified for the sulfur content within the QTL intervals. The sulfate transporter 4.1 (Glyma.02G095500) underneath (qS-01) is involved in the transport of sulfate ions across cell membranes. The iron-sulfur cluster assembly protein (Gm05_3273418) is involved in electron transfer reactions and other redox processes.

It is clear, based on the above discussion, and except of seed N, P, and S QTL, detected on Chr 16 [23], no QTL of seed N, P, and S were previously detected in mature seed for the accumulation of these nutrients. Therefore, our research showed that 11 QTL reported here were not previously identified, therefore, they are novel.