Delineation of Genotype X Environment Interaction for Grain Yield in Spring Barley under Untreated and Fungicide-Treated Environments

Barley (Hordeul vulgare L.) is the fourth most important cereal crop based on production and cultivated area. Biotic stresses, especially fungal diseases in barley, are devastating, incurring high possibilities of absolute yield loss. Identifying superior and stable yielding genotypes is crucial for accompanying the increasing barley demand. However, the identification and recommendation of superior genotypes is challenging due to the interaction between genotype and environment. Hence, the present investigation was aimed at evaluating the grain yield of different sets of spring barley genotypes when undergoing one of two treatments (no treatment and fungicide treatment) laid out in an alpha lattice design in six to seven locations for five years, through additive main effects and multiplicative interaction (AMMI), GGE biplot (genotype + genotype X environment), and stability analysis. The combined analysis of variance indicated that the environment was the main factor that contributed to the variation in grain yield, followed by genotype X environment interaction (GEI) effects and genotypic effects. Ten mega environments (MEs) with five MEs from each of the treatments harboured well-adapted, stable yielding genotypes. Exploiting the stable yielding genotypes with discreet use of the representative and discriminative environments identified in the present study could aid in breeding for the improvement of grain yield in spring barley genotypes.


Introduction
Barley (Hordeum vulgare L.) is one of the most widely grown cereal crops based on cultivated area and production quantity. It is the fourth most popular cereal (146 × 10 6 tonnes) after wheat (771 × 10 6 ), rice (787 × 10 6 ) and maize (1210 × 10 6 ) [1], supplementing the world's food and fodder requirements, alongside its utilization in the beer industry as raw material [2,3]. In the next five decades, the deployment of coarse grains as feed in developing countries is expected to increase, accounting for 56% of food grain demand [4], leading to increased production pressure on the cultivation of barley. To meet the increasing global food demand, the world barley production needs to be augmented by 54% in the next five decades [5]. World barley production has reached 158 × 10 6 tonnes, with 1.5 × 10 6 tonnes produced in Sweden [1]. Assuring food security through the evaluation, identification and development of high-yield varieties is one of the core objectives of the plant breeding program. Grain yield is a complex quantitative trait influenced by genetic and environmental factors [6,7]. However, problems arise in recommending a genotype with high yield due to the complex nature of grain yield and interactions between genetic, environmental, edaphic factors. Among these issues, genotype (G) X environment (E) interaction (GEI) is one of the major obstacles in exploiting and gaining full advantage of the genetic potential of genotypes, thereby slowing the progress of breeding [8]. The existence of GEI in cultivars can be confirmed based on noticeable disparity in the phenotypic performance of the genotypes in different environmental conditions, which arises due to variation in the genetic potential of genotypes and their ability to adapt for different environmental conditions [9]. Hence, the existence of GEI in crops will decrease the association between genotype and phenotype, leading to ambiguity in the selection and recommendation of genotypes to specific environments or locations [10]. The reduced selection efficiency of superior genotypes due to GEI could be conquered by evaluating genotypes in multiple locations/environments with the aim to identify stable, environment-specific genotypes [9,11] and attaining more stable and higher yields. The different statistical methodologies employed in dissecting the role of GEI to identify desirable genotypes in multiple environmental trials can be categorized into two types: univariate and multivariate methods. Out of all the available methods of depicting GEI, additive main effect multiplicative interaction (AMMI) and genotype + genotype × environment interaction effect (GGE) models are extensively used for their ability to detect GEI through genotype ranking across environments [12].
Achieving the targeted yield improvement is decelerated due to crop losses associated with various intrinsic and extrinsic factors, of which, diseases alone can cause crop losses of up to 20% of global production. Among all the diseases, fungal diseases have attained special attention due to their widespread nature and their ability to influence yields by anywhere from 1% to 100%, depending on the pathogen strain and host resistance to infection [13]. Disease management relies on the choices made regarding crop rotation, tillage, cultivars used, and the use of fungicides [14,15]. The quickest and most reliable measure in disease control for ensuring good yield is employing fungicides. However, the response to disease control practices such as the use of fungicides on cultivars with diverse genetic backgrounds can vary, due to their variation in sensitivity to environmental conditions and adaptation ability in different microclimate environments [16,17]. However, the increased application of fungicides in crop production is unsustainable due to the increased production costs and bio-augmentation through environmental contamination [18][19][20][21]. Hence, the identification and development of cultivars with high/stable yield across different environments with the marginal requirement of fungicides is desirable and will favour sustainable barley production. Understanding genotypic interaction with fungicide application will open avenues to lower crop production costs through limited fungicide application. da Silva et al. [21] studied the effect of fungicide-treated and untreated conditions on the yield of Brazilian oat cultivars (Avena sativa L.) and identified environment-specific genotypes with adaptability and stability. However, there are only a few reports on grain yield with/without fungicide application and the adaptability to different environments in barley [22][23][24]. Hence, the current investigation was aimed at studying the effect on grain yield in response to fungicide application and the identification of stable genotypes, better adapted to different locations in Sweden through AMMI and stability indices under untreated and fungicide-treated conditions.

Mean Genotypic Performance
The meteorological characteristics showed wide variation in temperature, humidity and precipitation across all five years of evaluation (Table S1). The mean grain yield of genotypes varied widely, indicating substantial variation in the genotypic potential of the genotypes under evaluation (Table 1). The genotypes with the highest mean grain yields were G44 with 0.882 kg m −2 and 0.945 kg m −2 in Y1, G3 and G4 with 0.959 kg m −2 and 1.070 kg m −2 in Y2, G12 and G8 with 0.609 kg m −2 and 0.622 kg m −2 in Y3, G34 and G38 with 0.808 kg m −2 and 0.898 kg m −2 in Y4, and G32 and G3 with 0.897 kg m −2 and 0.950 kg m −2 in Y5 under untreated and treated conditions, respectively ( Figure 1). The application of fungicide significantly improved the mean grain yield, and this increase was highest in Y4, followed by Y2, Y1 and Y5 ( Figure 2).    the grain yield (X axis represents year and evaluated environments. Y axis repres evaluated).  Year of evaluation

Genotypic Potential and Stability Indices
A total of 24 genotypes in Y1, 14 genotypes in Y2, 18 genotypes in Y3, 24 genotypes in Y4 and 16 genotypes in Y5 had positive genotypic potential (GP) under both untreated and treated conditions. The genotypic potential (GP) values, AMMI stability values (ASV) and genotypic selection index (GSI) values across all evaluated years are presented in Table S2. The average environmental coordinate (AEC) axis of the biplot recommended the stable genotypes under untreated (eight in Y1, eleven in Y2, fourteen in Y3, twenty in Y4, and eight in Y5) and treated conditions (fourteen in Y1, seven in Y2, twenty-two in Y3, ten in Y4, and six in Y5) ( Figure 5). Stable genotypes, according to AEC with positive GP values accompanied by superior stability, were identified under untreated and treated conditions (Table S3), with grain yield ranging from 0.571 to 0.959 kg m −2 and from 0.576 to 1.035 kg m −2 under untreated and treated conditions, respectively ( Table 4).

Discussion
Barley breeding with a focus on developing high-yield and admissibly stable genotypes is challenged by the varied performance of the genotypes under different locations or environments. Genotype evaluation in multiple environments and the identification of the best performing genotype lacks efficienct selection due to the interaction of genotypes with the environment, thereby reducing the correlation between the phenotype and genotype, leading to ambiguity in identifying the best performing genotype. Along with GEI, fungal diseases are one of the major problems facing barley cultivation, causing substantial yield losses, which could be managed through fungicide administration. Moreover, genotypic interaction with the fungicides, along with the confounding effect of GEI, toughens the process of discerning promising genotypes. Therefore, the present investigation was undertaken to identify high-yield and stable barley genotypes under untreated and fungicide-treated conditions using AMMI-GGE biplot analysis, which could aid in the reduced usage of fungicides, thus increasing sustainable production. Employing AMMI and GGE biplot approaches in understanding GEI is considered to be a systematic approach for grouping the genotypes in accordance with the environment through ranking based on the phenotypic performance and for understanding the relationships between the tested genotypes and environments [25][26][27][28]. The results of this experiment revealed the complex nature of grain yield and the confounding effects of fungicides, such as significant improvement of the mean grain yield in all tested years, except Y3. The preliminary economic analysis suggests that fungicide spraying resulted in a 4.5% increase in malting barley profits with substantial yield improvement [22]. However, fungicide application always does not translate into yield improvement, which could be explained by the variation in the magnitude of disease influence on some genotypes [22,29]. In the current investigation, the non-significant differences in yield recorded in Y3 might be associated with the low humidity due to diminished rainfall, making it unfavourable to disease incidence. The genotypes evaluated under the current study revealed significant differences in grain yield across all years, indicating the existence of genetic differences in yield under untreated and fungicide-treated conditions. The results of the analysis of variance from AMMI indicated that a major portion of the sum of squares of grain yield under untreated and treated conditions can be attributed to location, followed by GEI and genotype (Table 2). In the current investigation, location accounted for the largest share of sum of squares, indicating the diverse nature of environments and that a major part of grain yield variation was due to variation in location. Similar findings have been reported previously [9,[30][31][32]. The application of the AMMI model for the decomposition of GEI effects revealed that the combination of IPCA1 and IPCA2 together explained 55.9-69.3% and 55.6-73.0% of total GEI under untreated and treated conditions, respectively, and the scores of IPCA1 and IPCA2 revealed 32 and 29 stable genotypes across all the years in untreated and treated conditions, respectively ( Figure 6). IPCA1 and IPCA2 scores are a depiction of the genotypic stability across the environments; genotypes with low scores are expected to have high stability across all the tested environments. The use of both IPCA1 and IPCA2 is a strong approach for the identification of stable genotypes since it allows for conclusions about consistency in genotypic performance and their divergence, along with the role of the environment [33].
The angle between the environment vectors conveys the association among the evaluated environments [34][35][36]. In the current investigation, the angle between the environments in untreated (three to five environments/year) and treated (three to six environments/year) conditions was less than 90 • , inferring a positive association among the environments. Delineation of the evaluated environments into groups based on the cosine of the environmental vector angle has been reported previously in barley [28,31,37,38]. Among all the tested environments, E3 in Y1, E2 in Y2, E1 and E5 in Y3, E4 and E5 in Y4, and E5 in Y5 were highly discriminative and representative environments under untreated conditions. Under treated conditions, E2 and E5 in Y1, E6 in Y2, E5 and E6 in Y3, and E5 in Y4 were highly discriminative and representative. The test environment efficiency is evalu-ated based on discrimination and representation ability [37]. The discrimination ability of an environment is revealed by the length of environmental vectors, where the length of each vector is directly proportional to the standard deviation of the environment itself [25]. In the present investigation, highly discriminative environments with good representativeness under untreated and treated conditions were the candidates for delineating the broadly adapted genotypes, while the discriminative and non-representative environments identified were better suited to studying genotypes with special adaptability [39]. Among all the tested genotypes in the present study, 50% to 61% of genotypes represented the positive genotypic potential for grain yield under untreated and treated conditions in each year, indicating their superior performance with respect to grain yield. Ndiaye et al. [40] used the genotypic potential index to identify the better performing sorghum genotypes with respect to grain yield and biomass. Among the tested genotypes, three to twelve genotypes in every evaluated year and treatment had a smaller perpendicular line to the AEC axis of the biplot (Figure 6), inferring the stability of genotypes. Similar results were reported by Kendal et al. [28] in barley. Based on the AEC, ASV and GSI indices, twelve genotypes in Y1, seven genotypes in Y2, fifteen genotypes in Y3, nineteen genotypes in Y4 and eight genotypes in Y5 were identified as demonstrating stable performance under all tested environments ( Table 4). The ASV indicates the stable genotypes (with ASV values near to 0 indicating stability) based on the balanced measures from the sum of square values of IPCA1 and IPCA2, whereas GSI index integrates the ASV with the grain yield of the genotypes, thereby further increasing the selection efficiency for better genotypes. ASVs are commonly used in studies for the identification of stable barley genotypes under multiple environmental studies [7,11,41]. Among the common stable genotypes, G53 (Y1), G3 (Y2), G8 (Y3), G38 (Y4) and G33 (Y5) under untreated conditions and G16 (Y1), G44 (Y3), G43 (Y4) and G28 (Y5) under treated conditions manifested lower IPCA1 and IPCA2 values along with higher yields, indicating their stability across the evaluated environments. Similar results were reported by Elakhdar et al. [42] in barley under salt stress conditions. Which-won-where analysis of the biplot identified mega environments comprising of three to five locations in each evaluated year (Figure 7), allowing breeders to identify good test environments for the detection of genotypes adapted for the specific environmental factors [39,43,44]. In the present investigation, within each year, locations were partitioned into different MEs, and the pattern of grouping was different between untreated and treated conditions, with one to three common environments between the untreated and treated conditions, which infers that these common environments are suitable for assessing the adapted genotypes under both untreated and fungicide-treated level evaluations. WWW analysis of biplots is the most efficient way of delineating the GEI of genotypes through plotting the multi-location data of environments and genotypes in a polygon view of a GGE biplot [45]. In the present study, WWW plots revealed that G44 (Y1 untreated and treated), G20 (Y2 untreated), G19 (Y3 untreated), G8 (Y3 treated), G48 (Y4 untreated), G28 (Y4 treated), G3 (Y5 untreated) and G32 (Y5 treated) were the vertex genotypes, with higher yields in each ME. WWW plots of GGE biplots is an efficient method of determining the best genotypes in mega environments [42]. The superior-yield, winning genotypes identified in the MEs could be considered as checks in fungicide evaluation trials within the evaluated environments [46]. Because of the significant contribution of location to the variation in grain yield, the ideal genotype identified in multi-environment evaluation should have high performance, combined with stability across environments. In the present investigation, genotypes suitable for multiple environments with stable performance were identified under both untreated and treated conditions in Y1 (G49 and G53 in untreated condition, and G10, G16, G21, G36, G37, G44 and G48 in treated condition), Y2 (G3, G29 and G38 in untreated condition), Y3 (G3 in untreated condition and G10, G17, G26, G46 and G50 in treated condition), Y4 (G7, G19, G21, G38, G40, G42, G46 and G47 in untreated condition and G17 in treated condition) and Y5 (G21, G22 and G28 in treated condition). Similarly, da Silva et al. identified stable and better-adapted oat genotypes for yield and grain quality under untreated and fungicide-treated conditions [21]. Vaezi et al. evaluated barley genotypes for three years and identified stable genotypes based on stability statistics and the GGE biplot approach [9].
al. identified stable and better-adapted oat genotypes for yield and grain quality under untreated and fungicide-treated conditions [21]. Vaezi et al. evaluated barley genotypes for three years and identified stable genotypes based on stability statistics and the GGE biplot approach [9].

Experimental Site and Plant Material
The study was executed with different sets of spring barley genotypes in seven locations over five years (2016-2020) under the Sweden National Trails program with two different treatments (untreated and treated with fungicide). A diverse set of spring barley genotypes were evaluated each year at seven locations and a new set of genotypes were used each year, as per the updated list of released/popularly cultivated genotypes (Table  S4). The five different years under genotypic evaluation were denoted as Y1 (2016), Y2 (2017), Y3 (2018), Y4 (2019) and Y5 (2020). In each year, genotypes were evaluated at six to seven different locations in Sweden (Figure 7). Location descriptions along with meteorological data of each environment are presented in Table S1, and the meteorological data were obtained from https://sverigeforsoken.se/s (Accessed on 03 September 2021). All the spring barley genotypes were raised in alpha design with two replications in each environment, using standard agronomic practices except for fungicide application. Treatment was imposed by the application of an extra dose of fungicide to determine the variety of fungal resistance. During Y1, Flexity (Flexity ® , from BASF), Proline EC250 (Proline EC250, from BAYER) and Comet Pro (Comet ® Pro, from BASF) were applied for the treatment plot in all locations, and during Y2, Flexity, Siltra Xpro (Siltra Xpro from BAYER), Comet Pro was applied. In Y3, Y4 and Y5, Talius (Talius ® from Corteva Agriscience), Siltra Xpro and Comet pro were applied to maintain treatment. During physiological maturity,

Experimental Site and Plant Material
The study was executed with different sets of spring barley genotypes in seven locations over five years (2016-2020) under the Sweden National Trails program with two different treatments (untreated and treated with fungicide). A diverse set of spring barley genotypes were evaluated each year at seven locations and a new set of genotypes were used each year, as per the updated list of released/popularly cultivated genotypes (Table S4). The five different years under genotypic evaluation were denoted as Y1 (2016), Y2 (2017), Y3 (2018), Y4 (2019) and Y5 (2020). In each year, genotypes were evaluated at six to seven different locations in Sweden (Figure 7). Location descriptions along with meteorological data of each environment are presented in Table S1, and the meteorological data were obtained from https://sverigeforsoken.se/s (Accessed on 03 September 2021). All the spring barley genotypes were raised in alpha design with two replications in each environment, using standard agronomic practices except for fungicide application. Treatment was imposed by the application of an extra dose of fungicide to determine the variety of fungal resistance. During Y1, Flexity (Flexity ® , from BASF), Proline EC250 (Proline EC250, from BAYER) and Comet Pro (Comet ® Pro, from BASF) were applied for the treatment plot in all locations, and during Y2, Flexity, Siltra Xpro (Siltra Xpro from BAYER), Comet Pro was applied. In Y3, Y4 and Y5, Talius (Talius ® from Corteva Agriscience), Siltra Xpro and Comet pro were applied to maintain treatment. During physiological maturity, the crop was harvested for grain yield, and the data were reported as grain yield kg/square meter (kg m −2 ).

Statistical Analysis
The grain yield data of the spring barley genotypes were assessed for stability and G X E interaction using the AMMI model with GGE biplots under control and elevated fungicide treatment environments using GEA-R (GEA-R, CIMMYT, Mexico) [47]. The AMMI analysis has been found to be reliable in capturing a large proportion of G X E sum of squares, which clearly separates the main effects and interaction effects. Hence, it is ordinarily the first choice model when both the main effects and interaction effects are important, which is the usual case with the yield trials [48,49].
GGE is a linear-bilinear model, which is recommended when the environments are the main source of variation in relation to the contributions of genotypes and GEI with respect to the total variability. At the same time, this technique allows the determination of mega-environments (GEA-R, CIMMYT, Mexico).
The model employed for AMMI and GGE analysis is given below, and the results of the analysis were presented in the form of biplots.
AMMI analysis: τ n Y in δ jn + ε ij GGE analysis: τ n Y in δ jn + ε ij Y ij represents the yield of the ith genotype in jth environment; grand mean, genotype and environment deviations from grand mean are represented by µ, g i and e j . τ n represents the eigenvalue of principal component (PC) analysis axis n. The number of PCs and error terms are denoted by N and ε ij .
Analysis of variance of the grain yield data was performed using open software R [50] with the agricolae package [51]. The genotypic potential (GP) index was calculated according to Ndiaye et al. [40] by employing the formula below. A genotype with a positive GP value indicates good genotypic potential and vice versa for a negative value.
Genotypic potential index = Y .
I j − Y Y Y ij represents grain yield of a given genotype i in a given environment j, while Y denotes overall mean grain yield.
The AMMI stability values (ASV) were calculated using the method formulated by Purchase et al. [52] via the following formula.
Ammi stability value (ASV) = SS IPCA 1 SS IPCA 2 (IPCA 1 ) where SS in the equation denotes the sum of squares of the first (IPCA1) and second (IPCA2) interaction principal components, and the genotypic scores are obtained from the AMMI model. The genotype selection indexes (GSIs) of the evaluated genotypes in the present investigation were calculated using the following formulae as obtained from Farshadfar and Sutka [53].
where GSI i refers to genotype selection index of the ith genotype; Y i refers to rank of mean grain yield of ith genotype and ASVi denotes the rank of ASV of ith genotype.

Conclusions
The present investigation revealed that the grain yields of barley genotypes are largely affected by location, followed by GEI and genotypes. Fungicide application significantly increased the grain yield and altered genotypic stability. Genotypes adapted to multiple environments manifesting stable yield were identified under untreated (G49 and G53 in Y1; G3, G29 and G38 in Y2; G3 in Y3; and G7, G19, G21, G38, G40, G42, G46 and G47 in Y4) and treated (G10, G16, G21, G36, G37, G44and G48 in Y1; G10, G17, G26, G46 and G50 in the treated condition in Y3; G17 in Y4; and G21, G22 and G28 in Y5) conditions, which are possible candidates for the molecular dissection and further yield improvement of spring barley in the targeted locations. The MEs and winning genotypes (G44 in Y1, G20 in Y2, G19 and G8 in Y3, G48 and G28 in Y4, and G3 and G32 in Y5) identified in the present study advocate precise testing of germplasm for grain yield under untreated and fungicide-treated trials. Prudent use of the identified genotypes from evaluation as pre-breeding material will hold potential in the development of barley genotypes with broad adaptation and stable yield.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/plants12040715/s1, Table S1: Details of meteorological characteristics of the environments under evaluation; Table S2: Details of IPCA scores of the genotypes evaluated across the environments and treatments; Table S3: Details of stability indices of the evalauted genotypes; Table S4; List of the genotype names evaluated in the present investivation.