Adaptability and stability of transgenic soybean lines and cultivars in the Brazilian macroregion 3 assessed by using parametric and nonparametric methods

This study aimed to evaluate the adaptability and stability of 20 soybean genotypes, ten of which were lines of Monsoy Ltd breeding program, whereas the other 10 genotypes were commercially used cultivars. The experiments were carried out in the agricultural years of 2005/2006 and 2006/2007, in Brazil, specifically in the soybean microregions 301, 302 and 303 (part of Minas Gerais, Goiás and São Paulo states). A randomized complete block design was used for all genotypes in each location, with three repetitions. Individual and joint analyses were done considering genotype yield in the different locations. Subsequently, genotypic adaptability and stability were evaluated by the methods of Eberhart and Russel (1966), Lin and Binns modified by Carneiro, Annicchiarico and Centroid. All methods presented partial coherence on classifying the best genotypes and allowed the identification of the transgenic lines L1 and L4, and the cultivars M-SOY 8199 RR, M-SOY 8045 RR, and Valiosa RR as the most promising ones to be grown in the microregion 3 because they have shown both stability and wide adaptation combined with outstanding grain yield. Lines L1 and L4, both with superior grain yield than the controls M-SOY 8199 RR and Valiosa RR, were classified as those with adaptability and stability to favorable environments. On the other hand, the lines L3, L10 and the cultivars M-SOY 8064, M-SOY 7908, and M-SOY 8045 RR were the most stable and productive genotypes for unfavorable environments.


INTRODUCTION
The soybean crop has been considerably developed in Brazil since the last three decades and in the last harvest season, it represents 49% of the total area aimed for grain production in the country.During the harvest season of 2014/2015, Brazil produced 95 million tons of soybean in an area of approximately 31 million hectares.
Currently, the country is the second largest producer and exporter of the crop in the world (CONAB, 2015).
This expansion and the increasing grain yield are mainly due to technological developments and correct management of the crop (MAPA, 2014).The increasing soybean yield in the last years is essentially due to genetic breeding and development of new technologies of production (Freitas, 2011).The crop's expansion has been promoted towards new agricultural fields as a result of the large adaptability of the crop to different environmental conditions.
The term environment can be understood, in this context, as a series of conditions in which the plants grow and develop themselves involving aspects such as location, season, year, cultural practices or the combination of them (Rocha, 2002).Throughout the breeding process, genetic materials are tested in a wide range of croplands for posterior performance evaluation and genetic superiority proof.In this context, the occurrence of genotypic and environmental interactions (GxE) is common, for instance, those interactions affect crop behavior by reflex of environmental oscillations (Cruz et al., 2012).Therefore, knowing and evaluating the elements that compose this interaction are very important for genetic breeding programs because they allow further identification of genotypic responsiveness and predictability as a result of environmental variations.
The adaptability of a certain cultivar is related to its capacity of taking advantageously use of environmental stimulation; the stability is the ability to show a behavior that is highly predictable as a matter of environmental modification (Cruz et al., 2012).
Adaptability studies using parametric methodologies such as Eberhart and Russel (1966), Centroid (Rocha et al., 2005) and nonparametric methods such as that of Lin and Binns (1988) modified by Carneiro (1998) and Annicchiarico (1992), have been largely used in soybean crop in order to assess the stability and adaptability of genotypes (Vasconcelos et al., 2010;Barros et al., 2010;Marques et al., 2011;Oliveira et al., 2012;Carvalho et al., 2013;Polizel et al., 2013).Likewise, scientific researches, such as the one carried out by Silva Filho et al. (2008), reported that both nonparametric methods used in this study are concordant and identify lines of great performance and wide stability.
Thereby, this study aimed to evaluate the soybean performance, adaptability and stability of 20 genotypes of Monsoy Ltd breeding program by parametric and nonparametric methodologies in the soybean producing macroregion 3.

MATERIALS AND METHODS
The experiments were done in Brazil, in the soybean producing macroregion 3, which covers part of the states of Minas Gerais, Goiás and São Paulo, specifically in the microregions 301, 302 and303 (Table 1) during the agricultural years of 2005/2006 and 2006/2007, aiming to assess the performance of different soybean lines.In Figure 1, it is shown that Brazil is subdivided into 5 macroregions and 29 soybean producing microregions.
The experiments were carried out in 8 municipalities of three different states: Goiás, Minas Gerais, and São Paulo, as shown in Table 1.It was evaluated 10 lines (L1 to L10) and developed by Monsanto Soybean Breeding Program -Monsoy Ltd, essentially from Morrinhos-GO research station, and 10 controls (commercial cultivars) (L11 to L20) of different maturity groups (Table 2).
All soybean lines assessed in the experiment were driven by the SPD method (single pod descendent), that is, starting from F2 to F5 generation, the procedure of picking one single pod per plant was performed.In the meantime, the seeds were sown from three to four rows of 5.0 m length, consisting of 12 to 15 seeds per linear meter, respectively.
With respect to field preparation, a burndown herbicide application for 14 days before sowing using the herbicide Roundup WG® was done with a dosage of 1.5 kg per hectare.Right before the sowing, complete soil analyses of all locations were done, as well as fertilizer applications according to the soil requirements and crop recommendations; the fertilizer formulation used was 2-28-20.
The sowing was done under no-till crop system using a plot seeder called Semeatos® SHP 249.During sowing process, an insecticide application was done at planting furrows with Cruizer® (300 g ha -1 ), and also, an inoculation with Gelfix® (10 doses ha -1 ) was done.
Roundup Ready® applications were performed 20 days after sowing on a dosage of 2.0 L ha -1 in order to control weeds.Meanwhile, insecticides and fungicides, registered for the crop, were sprayed as often as necessary.
The experimental design used was randomized complete blocks with three replications.Each experimental plot was formed by 4 soybean plant rows with 5 m length, spaced at 0.5 m within rows.The useful area was composed of 2 central lines, wherein the two external rows were discarded as borders, resulting in a useful area of 5 m 2 .
It was determined that the grain (kg ha -1 ) in the experimental plots were harvested through the use of a plot harvester Almaco® Company, model SPC-20.Whenever necessary, the soybean seeds were dried to a moisture content of 13% in a gas drier at Monsoy station, Morrinhos-GO.The seeds of each plot were kept in a cloth bag and weighed on a digital scale.
Data from grain yield were submitted to individual (each municipality separately) and joint variance analyses.The environmental variation source was composed of 3 soybean producing microregions with 9 municipalities and 2 agricultural years.Similar procedure was adopted by Oliveira et al. (2012).Before the joint analyses, the homogeneity of residual variance was checked by dividing the highest and smallest numbers of the mean square error.Since it was higher than 7, the degrees of freedom were adjusted.
After detecting the existence of GxE interaction, adaptability and stability analyses was conducted by using the methodology of Eberhart and Russel (1966), Lin and Binns (1988) modified by Carneiro (1998), Annicchiarico and Centroid.The statistical analyses were done using the Genes computer program (Cruz, 2013).

RESULTS AND DISCUSSION
The occurrence of GxE interaction (  genotypes in the sixteen studied environments, which represent the three soybean producing microregions.Due to the fact that the interaction genotype x cultivation area is significant, grain yield was influenced by either the genotype or the environment.Similar results were observed by Rocha and Velho (1999) while studying the same interaction (genotype x environment) for grain yield of soybean lines with different maturity groups.
The experimental coefficient of variation was of low magnitude (4.5%), indicating good experimental precision.Furthermore, the CV was lower than 16%, which is considered the maximum coefficient accepted for soybean grain yield according to Carvalho et al. (2003), and lower than what it was found in other studies (Barros et al., 2009(Barros et al., , 2010(Barros et al., , 2012;;Vasconcelos et al., 2010;Marques et al., 2011;Carvalho et al., 2013).
The existence of GxE interaction highlights differences on the behavior of genotypes in responsiveness to environmental fluctuations (Cruz et al., 2012), and therefore, justifies the study of adaptability and stability, allowing a better understanding regarding each genotype and future cultivar recommendations.
According to Barros et al. (2010), the GxE interaction event represents one of the main difficulties found by breeding programs, whether in cultivar selection or recommendation stages.In this context, it is undeniably important to know the adaptability and stability of genotypes to different growing regions in order to identify Adopting adaptability and stability analyses by using the methods of Annicchiarico (1992) (Table 4) and Eberhart and Russel (1966) (Table 5), it was possible to classify the localities according to classes, as shown in Table 4.It is noticed that from the sixteen classified environments, 50% were classified as favorable and the other 50% as unfavorable environments.
As stated in Eberhart and Russel (1966), the ideal genotype is the one that reveals B1 equal to a nonsignificant unit and regression deviation, and as a result, it is a genotype of wide adaptation and high predictability.Still, this method also facilitates the identification of genotypes adapted to unfavorable environments, B1<1, and to favorable ones B1>1.
In Table 5, according to Eberhart and Russel (1966) methodology, it was observed that the lines L6 and L9 and the cultivars M-SOY 8008 RR and M-SOY 8199 RR have shown wide adaptation.However, it has a low predictability for all deviation variances which were significant and R 2 had low magnitude, with the exception of M-SOY 8008 RR showing R 2 equals to 72%.The lines L1, L2, L4, L8, L10 and the cultivars M-SOY 8248 RR, M-SOY 8360 RR, M-SOY 8287 RR, Valiosa RR have shown adaptation to favorable environments although these adaptations were of a low predictability, as detected by the significant deviations.In contrast, only the genotypes L1, M-SOY 8360 RR, and Valiosa RR showed R 2 that is higher 70%.Meanwhile, the lines L3, L5, L7, L10 and the cultivars M-SOY 8064, M-SOY 8000 RR, M-SOY 7908 RR, and M-SOY 8045 RR demonstrated adaptation to unfavorable environments, once again, with low predictability and R 2 of low magnitude (Table 5).Previous studies on soybean lines and cultivars in the same microregion using similar cultivars such as M-SOY 8000 RR, M-SOY 8045 RR, M-SOY 8199 RR, and Valiosa RR have also shown that all sixteen genotypes evaluated presented significant regression deviations and predominance of low values for R 2 (Oliveira et al., 2012).
With respect to the methodology of Eberhart and Russel (1966), which is based on regression analyses and consideration of the values of R 2 , whether it is low, there is indication that the regression by itself does not explain properly the genotypic behavior against the environmental oscillations.Analyses on soybean lines and cultivars by Polizel et at. (2013) also suggest the predominance of low values for R 2 , similar to what was found in the current study.Additionally, having studied transgenic cultivars, Carvalho et al. (2013) has found that 63% of all genotypes analysed in their experiments were classified as having low predictability which is similar to the results achieved in this study, and accordingly, they classified the cultivars as having wide or specific adaptation to favorable and unfavorable environments.By the Lin and Binns (1988) methodology, modified by Carneiro (1998), the genotypic performance is estimated through the parameter (Pi), which is related to the distance between the genotype in analysis from the best genotype, so that the lower the value of Pi is, the higher the genotypic adaptability and stability will be.In Table 5, the lowest values of general Pi and high grain yield for the lines L1, L2, L3, L4, L10 and cultivar M-SOY 8199 RR were verified.
The lines L1, L2, L4, and L10 showed lower values of

Table 3 )
was verified by F test (P>0.01)for the trait, grain yield of all 20 *Corresponding author.E-mail: posagro@ufu.br.Author(s) agree that this article remains permanently open access under the terms of the Creative Commons Attribution License 4.0 International License

Table 1 .
Regions and municipalities where the transgenic soybean genotypes were grown over two harvest seasons, in three different microregions.
Figure 1.Brazil's map subdivided into macro and micro soybean producing regions (Source: Kaster and Farias, 2002).

Table 2 .
Cultivars and lines evaluated during two consecutive harvest seasons, 2005/2006 and 2006/2007, in the soybean producing macro region 3.

Table 3 .
Summary of joint variance analyses of grain yield (kg ha -1 ) of 20 soybean genotypes grown in sixteen environments, soybean producing region 3, during the harvest years of2005/2006 and 2006/2007.*Significant at 0.01 of probability by F test. *

Table 4 .
Grain yield of 20 soybean genotypes in each locality and environmental index by the method of Annicchiarico (1992).

Table 6 .
Adaptability and stability parameters of soybean genotypes during the harvest years of 2005/2006 and 2006/2007 in the soybean producing region 3, based on the methodology of Annicchiarico (1992).

Table 7 .
Adaptability and stability parameters of soybean genotypes during the harvest years of 2005/2006 and 2006/2007 in the soybean producing region 3, according to the Centroid methodology.
Classification: 1-Probability of belonging to the indicated class; Class I: general adaptability; Class II: specific adaptability to favorable environments; Class III: specific adaptability to unfavorable environments; Class IV: poorly adapted.