Effect of Genotype-by-Environment Interaction on Oil and Oleic Fatty Acid Contents of Cultivated Peanuts

: Twenty-seven genotypes of varieties and advanced breeding lines were grown in two locations in three years with three replications to estimate the effects of the genotype-by-environment interaction (G × E) on the oil and oleic fatty acid contents of cultivated peanuts. Oil and oleic fatty acid contents were quantiﬁed using NMR and GC, respectively. The tested lines were genotyped with functional SNP markers from the FAD2A and FAD2B genes using real-time PCR and classiﬁed into four genotypes. The results indicated that Alabama was the environment that better discriminated the test genotypes during the year 2012. Eight promising selected genotypes #12, #15, ARSOKR, Brantley, GaHO, M04-149, M04-48, and SunO97R showed wide adaptation and high-oleic acids of 83.02%, 81.32%, 82.03%, 81.15%, 79.21%, 80.94%, 82.46%, and 82.18%, respectively. The Additive Main Effects and Multiplicative Interaction (AMMI) model that combines the conventional analyses of variance for additive main effects with the principal component analysis (PCA) for the non-additive residuals was applied to estimate the additive effects from FAD2A and FAD2B genes and the G × E interaction. The results indicated signiﬁcant G × E interactions for oleic fatty acid contents. No correlation between oil content and FAD2A and FAD2B genes was found. The FAD2B gene had a larger additive effect than the FAD2A gene. The results from this study may be useful not only for peanut breeders, but also for food processors and product consumers to select suitable cultivars.

Most of studies were conducted on breed peanut varieties with a high oleic acid content [9][10][11].For example, one high oleate cultivar, SunOleic 95R, includes 49% oil composed of 80.6% oleic acid and only 2.8% linoleic acid [11].In peanut, genes ahFAD2A and ahFAD2B, encoding enzyme fatty acid desaturase (FAD2), controls the quantity of oleic acid.The function of this enzyme is to convert oleic acid to linoleic acid in the fatty acid synthesis pathway, while the high-oleic mutants will appear if this enzyme is inactivated [12,13].In 1987, Norden et al. first identified high-oleic mutants F435-2-1 and F435-2-2, and the content of oleic acid was up to 80%, while the linoleic acid level was only 2% [14].The mutant ahFAD2A gene, with a G448A substitution, was located on the peanut A genome, and the ahFAD2B gene, with an insertion of one base pair in 442, was located on the peanut B genome [15][16][17].Subsequently, F435 has been used to develop a series of breeding lines with a high-oleic-acid characteristic [11].Other high-oleate mutants were produced through mutagenesis [18,19].In addition, new mutation sites were also identified in ahFAD2B with a substitution of C301G from Arachis hypogaea [17] and in FAD2H with a substitution of C37T from Arachis veigae [20].
Oil and oleic acid contents in peanut seeds are complex quantitative traits that are controlled by more than one gene and are affected by the environment.Genotype, environment, and G × E interaction can affect the performance of the varieties.The environment allows the varieties to exhibit their best advantage.G × E analysis is the final step for the breeders to select the promising breeding materials [21].Among the statistical techniques used to analyze and interpret G × E data, the additive main effect and multiplicative interaction (AMMI) and genotype-by-environment interaction (GGE) are two most popular analysis methods [22].
Until now, no research on the genotype-by-environment interaction on oil and oleic acid contents in cultivated peanut has been conducted.And before this study, we did not know which gene had a higher additive effect on oleic acid, considering the two genes ahFAD2A and ahFAD2B.The objectives of this research were to (1) determine the variability in oil and oleic acid among germplasm accessions; (2) determine the effects of the genotype x environment interaction on the oil and oleic acid contents; and (3) determine the additive effects of genes ahFAD2A and ahFAD2B.

Genotyping FAD2A and FAD2B through Real-Time PCR
Genotyping was carried out using functional SNP markers from the FAD2A and FAD2B genes through real-time PCR according to the published method.Peanut seed slices (75-150 mg) were put into a 2 mL microcentrifuge tube, along with 600 µL of P1 buffer from the Omega-BioTek kit (Doraville, GA, USA) and two 3 mm tungsten carbide beads (Qiagen, Valencia, CA, USA) for the purpose of extracting DNA.After that, the tissue was ground up for three minutes at 30 Hz using a Retsch Mixer Mill 301 (Leeds, UK).Using a DyNA Quant 200 fluorometer from Hoefer Pharmacia Biotech (San Francisco, CA, USA), extracts were measured.To assess the amount and caliber of each extraction, all samples were also placed onto a 1% agarose gel using a Low DNA Mass Ladder from Invitrogen (Carlsbad, CA, USA).The samples were then diluted to 10 ng/µL in order to perform real-time PCR.1× TaqMan Genotyping Master Mix (Applied Biosystems), 0.16 µM forward primer, 0.16 µM reverse primer, 0.4 µM VIC probe, 0.3 µM 6FAM probe, and 0.4 ng/µL of DNA made up the 25 µL total volume of the PCR reaction.AmpliTaq Gold polymerase and ROX, a passive internal reference to account for signal variance between wells, are included in the TaqMan Genotyping Master Mix.Applied Biosystems' ABI StepOne real-time PCR equipment was used to conduct each PCR experiment.The cycling conditions were as follows: one cycle of 60 • C for 30 s, one cycle of 95 • C for 10 min, 50 cycles of 95 • C for 15 s and 62 • C for 1 min, and one final cycle of 60 • C for 30 s.

Oil and Fatty Acid Composition Measurement
Five medium-sized and healthy, naturally dried peanut seeds were crushed into fine powder in a small plastic bag with a hammer.Approximately 200 mg crushed seed powder was transferred into a small column and then pressed into a small pellet.Oil was quantified through nuclear magnetic resonance (NMR) analysis on a minispec seed analyzer (Bruker Optics Inc., Houston, TX, USA) according to the published report [23].Fatty acid methyl esters (FAMEs) were prepared from seeds through alkaline transmethylation, and fatty acid composition was determined using an Agilent 7890A gas chromatography (GC) equipped with a flame ionization detector (FID) and an autosampler.Sample preparation, GC operation, and data collection followed the standard methods routinely used by Wang's lab [23].Genotyping was carried out using functional SNP markers from the FAD2A and FAD2B genes through real-time PCR.

Statistical Analysis
A mixed-model analysis of variance (ANOVA) was performed using the GLM procedure of SAS (SAS, 2008, Online Doc ® 9.2., SAS Institute Inc., Cary, NC, USA) and means were separated using Duncan's multiple test procedure.Excel (2016) was used to perform variance analysis of variable components among 27 tested varieties.R4.0.3 [24] was used for AMMI and GGE analysis.Narrow-sense heritability (h 2 ) estimate for FAD2A gene can be estimated as h 2  yb is variance for FAD2B × year, σ 2 ab is variance for FAD2A × FAD2B, y is number of years, and b is degrees of freedom.Two-factor analysis of variance for estimates of two-gene model genetic effects was suggested by Cockerham [25] (Table 2).
Table 2. Two-factor analysis of variance for estimates of two-gene model genetic effects suggested by Cockerham (1963).

Analysis of Variance and Variability
A significant variability in oil and fatty acid composition among the 27 tested varieties was detected (Table 3, Dataset S1).Among all of these traits, the variance of C18:1 (162.64) and C18:2 (116.65) was larger than those of the other traits.The maximum (Max) of C18:1 was 83.02%, while the minimum (Min) of C18:1 was only 42.95%.The distribution of fatty acid composition among different genotypes is shown in Figure 1.Similarly, the Max and Min of C18:2 were 35.01%and 1.59%, respectively.C18:2 had the highest difference (22.02-fold difference) between the Max and Min.C20:0 and C18:0 also had large differences (3.03-fold and 2.46-fold) between the Max (2.18% and 4.60%) and Min (0.72% and 1.87%) than the other traits.According to the variance and standard deviation (SD), several of these traits such as C18:1 and C18:2 could be potentially improved through peanut breeding.

Variability in Traits among Accessions, across Locations and Years
The results of statistical analysis of the variable components and their interactions are listed in Table 4.There were significant differences in the variability among the tested varieties (G) for all of the 10 traits.Year also had significant effects on all of the nine fa y acid contents, but not for oil content.No significant differences between replicates for all investigated traits were detected.According to the F values determined through comparisons among the three individual components (year, replicate, genotype), genotype had a significant effect on all of the traits except for C26:0 and C24:0.Year had a larger effect on C26:0 (F value = 102.25 and 96.11, respectively) and C24:0 (F value = 149.87 and 81.69, respectively) than genotype.Among the two-factor interactions (Y × R, Y × G, and R × G), Y × R interaction effects on the variability in C26:0 and C20:0 were significant.Y × G effects and R × G effects on the variability in six traits were significant, but they were not significant for the variability for four other traits (oil content, C16:0, C18:2, C18:1).Interestingly, the three-factor interaction (Y × R × G) effect on the variability was similar to Y × G and R × G effects, which had significant effects on the variability in six traits, but not for four other traits (oil content, C16:0, C18:2, C18:1).

Variability in Traits among Accessions, across Locations and Years
The results of statistical analysis of the variable components and their interactions are listed in Table 4.There were significant differences in the variability among the tested varieties (G) for all of the 10 traits.Year also had significant effects on all of the nine fatty acid contents, but not for oil content.No significant differences between replicates for all investigated traits were detected.According to the F values determined through comparisons among the three individual components (year, replicate, genotype), genotype had a significant effect on all of the traits except for C26:0 and C24:0.Year had a larger effect on C26:0 (F value = 102.25 and 96.11, respectively) and C24:0 (F value = 149.87 and 81.69, respectively) than genotype.Among the two-factor interactions (Y × R, Y × G, and R × G), Y × R interaction effects on the variability in C26:0 and C20:0 were significant.Y × G effects and R × G effects on the variability in six traits were significant, but they were not significant for the variability for four other traits (oil content, C16:0, C18:2, C18:1).Interestingly, the three-factor interaction (Y × R × G) effect on the variability was similar to Y × G and R × G effects, which had significant effects on the variability in six traits, but not for four other traits (oil content, C16:0, C18:2, C18:1).

AMMI1 Biplot Display
The AMMI analysis was utilized to quantify the impact of the environment on the genotypes.Biplot analysis is a valuable tool for interpreting AMMI models.Two types of AMMI biplots were generated: the AMMI1 biplot, which displayed the interaction between the genotype mean and environments; and the AMMI2 biplot (GGE biplot), which showed scores for IPCA1 and IPCA2 [26].From Figure 3, it is evident that Env1 had the greatest main effects and was favorable for the performance of most genotypes.Conversely, Env3 exhibited lower main-effect values, indicating little interaction with genotypes.Env2 had a positive PC score with a high mean value.Genotypes #12, M04-48, and SunO97R were identified as well adapted to both Env2 and Env3, suggesting these two environments as suitable for these three genotypes.Genotypes F435, Florunner, Exp27-1516, Ga02C, ARSOKR, SunO97R, #12, M04-48, and GaHO displayed PC1 scores close to zero, while other genotypes demonstrated a below-average oleic acid content with negative PC scores or an above-average oleic acid content with positive PC scores.Genotypes #16-1 and F435 had a lower C18:1 content, whereas genotypes #12, M04-48, and SunO97R had a higher C18:1 content.Moreover, Env1 had a large negative PC1 score, which positively interacted with genotypes with negative PC1 scores like Olin, and negatively interacted with genotypes with positive PC1 scores.Finally, the AMMI1 biplot statistical model was employed to identify G × E interactions in peanut.Genotypes #10, #12, ARSOKR, AT3085RO, Brantley, Ga02C, GaHO, M04-48, and SunO97R were suitable for planting in Env2 and Env3, while genotypes #15, #16, F435HO, Fla-07, FR458, M04-149, M04-88, SunO93R, and WT4-121 were considered favorable environments for Env1.

AMMI1 Biplot Display
The AMMI analysis was utilized to quantify the impact of the environment on the genotypes.Biplot analysis is a valuable tool for interpreting AMMI models.Two types of AMMI biplots were generated: the AMMI1 biplot, which displayed the interaction between the genotype mean and environments; and the AMMI2 biplot (GGE biplot), which showed scores for IPCA1 and IPCA2 [26].From Figure 3, it is evident that Env1 had the greatest main effects and was favorable for the performance of most genotypes.Conversely, Env3 exhibited lower main-effect values, indicating li le interaction with genotypes.Env2 had a positive PC score with a high mean value.Genotypes #12, M04-48, and SunO97R were identified as well adapted to both Env2 and Env3, suggesting these two environments as suitable for these three genotypes.Genotypes F435, Florunner, Exp27-1516, Ga02C, ARSOKR, SunO97R, #12, M04-48, and GaHO displayed PC1 scores close to zero, while other genotypes demonstrated a below-average oleic acid content with negative PC scores or an above-average oleic acid content with positive PC scores.Genotypes #16-1 and F435 had a lower C18:1 content, whereas genotypes #12, M04-48, and SunO97R had a higher C18:1 content.Moreover, Env1 had a large negative PC1 score, which positively interacted with genotypes with negative PC1 scores like Olin, and negatively interacted with genotypes with positive PC1 scores.Finally, the AMMI1 biplot statistical model was employed to identify G × E interactions in peanut.Genotypes #10, #12, ARSOKR, AT3085RO, Brantley, Ga02C, GaHO, M04-48, and SunO97R were suitable for planting in Env2 and Env3, while genotypes #15, #16, F435HO, Fla-07, FR458, M04-149, M04-88, SunO93R, and WT4-121 were considered favorable environments for Env1.

AMMI2 Biplot Display
To assess the environment cluster differentiation, genotype-specific adaptation, and G × E interaction, a biplot illustrating the performance of 27 genotypes in three environments was generated (Figure 4).In Figure 4, the environments were categorized into three sections.Among them, Env3 exhibited short spokes and demonstrated weak interactive forces, whereas Env1 and Env2 displayed long spokes, indicating their discriminatory nature.In the AMMI 2 biplot, the genotype GaGreen was more responsive since it was more distant from the origin and suitable for Env2 and Env3.With the exception of #10, #14, #16-1, AT3085RO, F435N, GaGreen, and Olin, most genotypes were located near the origin, implying their lower sensitivity to environmental interactive forces.Overall, based on the findings from AMMI1 and AMMI2, genotypes #12, #15, ARSOKR, Brantley, GaHO, M04-149, M04-48, and SunO97R were identified as the best performers in terms of high oleic contents, making them favorable for Env3.

AMMI2 Biplot Display
To assess the environment cluster differentiation, genotype-specific adaptation, and G × E interaction, a biplot illustrating the performance of 27 genotypes in three environments was generated (Figure 4).In Figure 4, the environments were categorized into three sections.Among them, Env3 exhibited short spokes and demonstrated weak interactive forces, whereas Env1 and Env2 displayed long spokes, indicating their discriminatory nature.In the AMMI 2 biplot, the genotype GaGreen was more responsive since it was more distant from the origin and suitable for Env2 and Env3.With the exception of #10, #14, #16-1, AT3085RO, F435N, GaGreen, and Olin, most genotypes were located near the origin, implying their lower sensitivity to environmental interactive forces.Overall, based on the findings from AMMI1 and AMMI2, genotypes #12, #15, ARSOKR, Brantley, GaHO, M04-149, M04-48, and SunO97R were identified as the best performers in terms of high oleic contents, making them favorable for Env3.

Discussion
A correlation between oleic acid content and FAD2A and FAD2B genes has clearly been demonstrated by many researches; however, the relative contribution of FAD2 genes (FAD2A and FAD2B) to oleic acid quality traits in peanut is still unknown [27][28][29][30].The mutant allele FAD2A is widely available in the U.S. peanut germplasm collection but the mutant allele FAD2B is only present in the selected genotypes such as SunOleic 95R and SunOleic 97R [31,32].There are no studies on estimating the contribution of these two mutant alleles to oleic acid and we do not know which of the mutant alleles could produce more oleic acid [32].Researchers have just examined how the interactions of additive alleles determine the content of oleic acid; however, each allele can have an additive effect as well.The additive effects of two alleles were calculated for improving the further understanding of the genetic control for oleic acid synthesis in peanut.
The AMMI analysis is a most popular method to study G×E interactions and it was used to quantify the effect of G, E, and G × E interactions on drought-related traits in peanut [33].In addition, the effectiveness of the AMMI procedure has been clearly

Discussion
A correlation between oleic acid content and FAD2A and FAD2B genes has clearly been demonstrated by many researches; however, the relative contribution of FAD2 genes (FAD2A and FAD2B) to oleic acid quality traits in peanut is still unknown [27][28][29][30].The mutant allele FAD2A is widely available in the U.S. peanut germplasm collection but the mutant allele FAD2B is only present in the selected genotypes such as SunOleic 95R and SunOleic 97R [31,32].There are no studies on estimating the contribution of these two mutant alleles to oleic acid and we do not know which of the mutant alleles could produce more oleic acid [32].Researchers have just examined how the interactions of additive alleles determine the content of oleic acid; however, each allele can have an additive effect as well.The additive effects of two alleles were calculated for improving the further understanding of the genetic control for oleic acid synthesis in peanut.
The AMMI analysis is a most popular method to study G×E interactions and it was used to quantify the effect of G, E, and G × E interactions on drought-related traits in peanut [33].In addition, the effectiveness of the AMMI procedure has been clearly demonstrated in other crops, such as wheat [34], soybean [35], maize [36], pear millet [37], and field pea [38].However, it still has its own limits; for example, it does not provide a measure for quantitative stability [39].In this study, 27 peanut genotypes were evaluated in a three-year (2010-2012) field experiment, which was conducted at two locations.The combination of the AMMI model and biplot made it possible to describe the genotype-byenvironment interactions effect more accurately.
Compared to regular peanuts, the high-oleic peanuts are more naturally resistant to oxidation because it is higher in monounsaturated fats.The peanut single kernel oleic acid distributions were influenced by seed size, seed maturity, growing environment, and season flower termination [40].The main chemical technique is gas chromatography (GC), which calls for 100 extractions or injections of a standard sample and an experienced operator.Although more affordable and quicker than the GC technique, refractive index approaches still need a significant amount of time to complete [41].However, they exhibit a good correlation with the principal GC method.Numerous breeding efforts are using near-infrared (NIR) techniques to measure this fatty acid chemistry.It is also critical to understand how costly and time-consuming it is to measure this chemistry.Therefore, it is better to select varieties by genotype with higher additive effects and it is easy to obtain the genetic gain.For companies, quality is important, so they need to consider where to grow these high-oleic peanuts.The results from this study could help breeders and companies to obtain peanuts with stable high oleic contents.

Conclusions
In the present study, high-oleic and stable genotypes, such as genotypes #12, #15, ARSOKR, Brantley, GaHO, M04-149, M04-48, and SunO97R, could be used as new potential genetic resources for improving the peanut varieties with contents of oleic.In addition, the results also indicated that the FAD2B gene had a larger additive effect than the FAD2A gene, which provides important values for breeding high-oleic peanut varieties by editing gene FAD2B.

Figure 1 .
Figure 1.The distribution of fa y acid composition among different genotypes.The unit of the Yaxis is "%".

Figure 1 .
Figure 1.The distribution of fatty acid composition among different genotypes.The unit of the Y-axis is "%".

Figure 2 .
Figure 2. The changes in fa y acid composition with different genotypes of FAD2 alleles.The unit of the Y-axis is "%".

Figure 2 .
Figure 2. The changes in fatty acid composition with different genotypes of FAD2 alleles.The unit of the Y-axis is "%".

Table 1 .
The genotypes of 27 accessions used in this study.

Table 3 .
Variability in oil percentage on dry weight and fatty acid composition on oil content among selected peanut accessions.

Table 3 .
Variability in oil percentage on dry weight and fa y acid composition on oil content among selected peanut accessions.

Table 4 .
Statistical analysis of variable components among 27 tested varieties.

Table 5 .
Comparison of ten investigated traits among 27 tested varieties: oil content is expressed as % of dry weight, and fatty acids as % of oil content.

Table 6 .
Duncan test on oleic fatty acid contents of 4 genotypes of Fad genes.

Table 7 .
Pearson correlation coefficients and probability values for seed oil content and fa y acid composition among 27 tested varieties.

Table 7 .
Pearson correlation coefficients and probability values for seed oil content and fatty acid composition among 27 tested varieties.