Genetic evaluation of Pinus taeda clones from somatic embryogenesis and their genotype x environment interaction

The objective of this study was to evaluate the genotype x environment interaction and to estimate the genetic components of variance and mean using mixed models in early selection of 238 clones of Pinus taeda propagated by somatic embryogenesis. The experiment consisted of a complete blocks design, with 12 replications, with one plant per plot, in four environments, at 1, 3, and 4 years of age. Estimates of heritability and of genetic gains in the evaluated environments showed good prospects for selection of superior genotypes. The effect of genotype x environment interaction was pronounced for all traits investigated. With the simultaneous selection for stability and adaptability, 10% genetic gain was obtained in relation to the mean of the commercial controls. This estimated gain indicates that the somatic embryogenesis technique has been effective in propagation of clones with good productive potential.


INTRODUCTION
In Brazil, Pinus taeda has presented better development in the South and Southeast regions (Martinez et al. 2012).The increase in yield observed in Pinus taeda plantations is mainly due to the use of genetically superior material derived from breeding programs (Mckeand et al. 2006, Martinez et al. 2012).In view of the positive impacts of Pinus breeding programs on the production of raw material suitable for the manufacturing of long fiber cellulose (Mckeand et al. 2006, Martinez et al. 2012), their implementation in agricultural corporations is fundamental for yield increase.Considering the limitations of genetic gains in programs traditionally developed by seminiferous propagation, cloning tends to play important role in the consolidation of the competence of Brazilian industries in this market.
The negative effects of ontogeny have led to difficulties in clonal propagation and have consequently made the use of Pinus taeda clones on a commercial scale unviable (Pullman and Bucalo 2011).The cuttings collected from adult Pinus taeda trees are difficult to root (Alcantara et al. 2007(Alcantara et al. , 2008)), and the production of clonal seedlings by the minicutting technique has also presented low rooting percentages (Alcantara et al. 2007, Andrejow andHiga 2009).As an alternative, somatic embryogenesis has been developed and used in Pinus PC Dias et al. taeda cloning programs (Pullman et al. 2006, Alcantara et al. 2008, Andrejow et al. 2009, Pullman and Bucalo 2011).This way, plant yield can be significantly improved due to the multiplication of desirable genotypes derived from breeding programs (McKeand et al. 2008).Thus, somatic embryogenesis should be used in breeding programs of Pinus taeda as long as the genotype x environment interaction comes from immature zygotic embryos, given the effects related to ontogenetic age in Pinus taeda (Pullman et al. 2006, Pullman et al. 2011).
In a forestry-breeding program, genetic evaluation of individuals and their relations with the planting environments is a fundamental step.Due to environmental variations, phenotype variations also occur in function of the genotype x environment interaction, being one of the greatest problems of breeding programs of any species, whether at the stages of selection or recommendation of cultivars.Nevertheless, analyses of phenotypic adaptability and stability can be used, which identify cultivars responsive to environmental variations and with foreseeable behavior (Cruz et al. 2004, Resende 2007, Rosado et al. 2012).In this context, the mixed models method (REML/BLUP) is considered as more accurate (Resende 2007), since it provides better experimental accuracy, and is more efficient than analysis of variance, especially in cases with unbalanced data.Moreover, the predicted genetic values can be used to estimate the adaptability and stability of genotypes using the harmonic mean of the relative performance of genetic values (HMRPGV), which allows estimating adaptability and stability simultaneously in a single parameter (Resende 2007).
In general, forestry-breeding programs consider the results of juvenile-mature correlation analyses to carry out early selection, due to the long crop rotation cycle.In the case of Pinus taeda, studies have presented good results with early selection (Paludzyszyn Filho et al. 2001, 2002, 2003, Isik et al. 2005, Mckeand et al. 2006, Martinez et al. 2012).The objective of this study was to evaluate the genotype x environment interaction and to estimate the genetic components of variance and mean using the mixed models (REML/BLUP) in early selection of Pinus taeda individuals propagated by somatic embryogenesis.

MATERIAL AND METHODS
The study was carried out by genetic-statistical analysis of part of the experimental network of Pinus taeda of the Klabin S.A. Corporation, which is composed of 238 clones propagated by somatic embryogenesis, using megapethophytes from immature seeds of matrices selected in the company.Somatic embryos were obtained by the methodology described in the U.S. Pat.N. 5506136 A (BECWAR et al., 1996).Clonal tests were set up in the states of Parana and Santa Catarina, in 2007, using seedlings at 10 months of age (propagated by somatic embryogenesis).Seedlings were produced in 55 cm³ tubes, using decomposed pine bark as substrate, with periodic fertilizations of NPK and micronutrients solution.Subsoiling was carried out at 50 cm depth.In the field, weed control was performed with herbicide (glyphosate) in the total area, one month before planting, and 4, 12, 18, 24 and 36 months after planting.Leaf-cutting ants control was carried out using formicide baits.The experimental consisted of a complete blocks design, with twelve replications, spaced 3 m x 2 m between plants, with one plant per plot, in four environments, two in Santa Catarina and two in Parana.Three lots of commercial seeds were used as controls.
According to the Köppen climate classification, environments 1 and 2, in the state of Santa Catarina, are characterized as Cfb; and environments 3 and 4, in the state of Parana, are located in a transitional climate region between Cfa and Cfb.Environments 1 and 2 have lower average temperatures and a greater number of frosts than environments 3 and 4. The soil of environment 1 is classified as Inceptisol, with clayey texture, and slightly rolling to rolling relief.The soil of environment 2 is classified as Oxisol, with clayey texture, and slightly rolling to rolling relief.Finally, the soil of environment 3 is classified as Ochrept or Umbrept, with medium texture, with rolling to steeply rolling relief.Environment 4 is classified as Psamment, with sandy and medium light texture, and rolling to steeply rolling relief.
Diameter -DBH (in cm, measured at 1.30 m from the soil surface), total height -Ht (in m), volume -Vol (m 3 ), and survival rate at 1, 3, and 4 years of age of Pinus taeda clones were measured.DBH was measured using a diameter tape, and height was obtained using a relascope.For volume calculation, the following formula was used: Vol = ( 3.1416 × DBH 2 4 ) × Ht × 0.5.
Survival rate was evaluated by counting the number of live trees per clone in the experiment at the time of measurements of DBH and Ht (at 1, 3, and 4 years of age).
Analyses were carried out by the estimate of variance components (Reml) and by the genetic value prediction (Blup), using the software Selegen-Reml/Blup (RESENDE 2002b).Variables were evaluated individually per environment, and in combination of environments.In evaluation of the individuals within each environment, the variables were analyzed using the univariate linear mixed model of the software Selegen-Reml/Blup, presented by Resende (2002a), according to the model: y = Xr + Zg + Wb + e, in which: y = data vector; r = replication effect vector (assumed as fixed) added to the overall mean; g = genotypic effect vector (assumed as random); b = block effect vector (assumed as random); and e = error or residue value (assumed as random).Uppercase letters represent the incidence matrices for the respective effects.The statistical model for the analysis of this experimental network in several environments, considering one observation per plot, is given by: y = Xr + Zg + Wb + Tge + e, in which: y = data vector; r = replication effect vector (assumed as fixed) added to the overall mean; g = genotypic effect vector (assumed as random); ge = genotype x environment interaction effect vector (assumed as random); b = block effect vector (assumed as random); and e = error or residue value (assumed as random).Uppercase letters represent the incidence matrices for the respective effects.
Stability and adaptability were simultaneously evaluated by the harmonic mean of relative performance of genetic values (HMRPGV), according to Resende (2007).All analyses were carried out using the software Selegen-Reml/Blup.With the predicted genetic values, genetic correlations were obtained between the traits evaluated in combined analysis with the environments.

Evaluation in each location
Considering the evaluations in the third and in the fourth years, heritability values for clones means in relation to the traits height, DBH, and volume were of high magnitude (from 60% to 82%), and significant by the likelihood ratio test at 5% significance.This fact results in high accuracies in the selection of clones propagated by somatic embryogenesis, indicating expressive genetic control for these traits in Pinus taeda clones (Table 1).These estimates are in agreement with those reported for Pinus taeda by Isik et al. (2003) for volume (0.70), and by Isik et al. (2005) for growth traits (0.50 to 0.75).Since the family structure is considerably different between these studies, it is inferred that growth traits in Pinus taeda are under moderate to strong genetic control, and that the somatic embryogenesis technique did not affect the expression of these traits.
The estimates of broad-sense individual heritability were lower than those obtained at the mean level of the clone, and varied according to the environment and year of evaluation (Table 1).Heritability estimates of low to moderate magnitude have been observed in other species propagated by somatic embryogenesis, such as in Pseudotsuga menziesii at five and a half years after planting (height = 0.25 ± 0.01; DBH = 0.21 ± 0.01; and volume = 0.20 ± 0.01) (Dean 2008); and in Picea glauca at four years after planting (height = 0.137 ± 0.041) (Wahid et al. 2012).
The lowest values for heritability, accuracy, and coefficient of genotypic variation at all the ages were observed in environment 1, in Santa Catarina (Table 1).The other environments presented better conditions for the development and expression of the genetic potential of clones, providing, in these cases, better conditions to detect existing variation and, consequently, greater possibilities of genetic gains with selection.Environment 1 presented edaphic traits inferior to those of the other environments, and this may have influenced gene expression of the clones propagated by somatic embryogenesis, which negatively reflected in the genetic parameters evaluated in this study.
The coefficient of genotypic variation (CV gi ) of the traits evaluated in this study had little variation, considering the three ages of study and the four environments.Environment 1 in Santa Catarina had the lowest coefficients of genotypic variation at the three ages of evaluation (ranging from 6.2 % to 10.5 % for height; 9.2% to 11.7 % for DBH; 0.6% to 5.5% for survival rate; and 21.8% to 28.8% for volume), as observed in Table 1.Of the traits evaluated, volume had the greatest coefficients of genotypic variation at all ages (greater than 20%).The presence of considerable genetic variability, as observed in this study, indicates the possibility of practicing selection among clones, especially for volume (Resende 2007).Thus, it is possible to obtain genetically significant gains in selection of Pinus taeda clones propagated by somatic embryogenesis.
No great variation was observed among the different ages for survival rate (

Combined analysis of environments
The heritability estimates reported in the combined analysis lead to expressive selective accuracies for the studied traits, especially for volume (Table 2).Nevertheless, these heritability estimates were low when compared with those found in the individual analysis per environment (Table 1).This indicates that individuals should be selected.The relatively low estimates for heritability in the traits evaluated in the combined analysis of environments (Table 2) suggest that other factors, besides genetics, strongly affect these traits, such as the environmental effects of sites and the genotype x environment interaction.
Corroborating the data obtained in this study, Xiang et al. (2003) observed in full-sib families of Pinus taeda that the ideal age for early selection, considering volume and the DBH, is from 4 to 5 years.Gwaze et al. (2001)  reported similar results when evaluating Pinus taeda families at 5 to 25 years of age.These results, once again, show that 1 year of age is not adequate for selection.This is because early selection does not reveal the presence of competition among plants, which is manifested in the evaluations at 3 and 4 years, in addition to the lower heritability.Significant genetic variability is observed by the likelihood ratio test at 5% significance among the clones evaluated in the combined analysis in the state of Santa Catarina, Parana, and in all environments, as shown by the heritability estimates and their standard deviations (Table 2).The values of the coefficient of genotypic variation (CV gi ) for DBH and height in the three years of study were of approximately 7%; however, considering the combined analysis among the environments, the volume showed values greater than 22% in the three years of evaluation.The coefficient of genotypic variation for volume demonstrates that selection of genotypes is possible; this is because the CVgi is greater than 10%, which is enough to practice effective selection among clones (Resende 2002).
Genotypic correlation among the environments (r gloc ) was moderate to high for almost all the traits evaluated in the first year of the study, ranging from 0.38 to 0.93.However, in the third and fourth years, genotypic correlation among the environments was of low to moderate magnitude for almost all the traits evaluated, ranging from 0.28 to 0.65 (Table 2).According to Table 2, the experiments revealed low coefficients of determination of the effects of the genotype x environment interaction at the ages evaluated.
Genotypic correlation greater than 0.67 are considered as high, and indicates that a single breeding program simultaneously and satisfactorily meets the demands of all the environments evaluated in the present work (Resende 2002a).In this study, the combined analysis of all environments was moderate, which requires differentiated selection for the different environments, indicating that some genotypes may have superior performance in one environment, PC Dias et al.
but not in another (Cruz et al. 2004, Resende 2007).In general, clones are more unstable than families; thus, there is a tendency of lower genotype x environment correlation in clonal tests.Nunes et al. (2002) reported that the response correlated by selection in an environment and gain in another environment has always been lower than the gain from direct selection in the environments when significant interaction is observed.
Genetic correlation among the environments indicates that selection of specific clones for each environment is recommended.In addition, from these results, adaptabilities and stabilities of clones should be taken into account when selecting these clones (Resende 2007).For analyses of genetic gain, stability, and adaptability in each environment, only the volume data will be discussed, since this trait tends to be the most representative at the initial stage of clone selection (Mckeand et al. 2006, Santos et al. 2006, Beltrame et al. 2012), and is of great commercial interest.

Genetic gains
Genetic gain in relation to the overall mean of the experiment, using the five best clones according to the genotypic values, was of approximately 50% in selection in the combined analysis of environments; greater than 69% in environment 1; 57% in environment 2; and greater than 100% in environments 3 and 4, in the state of Parana (Table 3).However, when compared with the mean value of the controls (matrices 161, 162, and 163), the genetic gain using the same five best clones decreased to values from 9 to 19% in the combined analysis of environments; 7 to 13% in environment 1; 8 to 18% in environment 2; 30 to 70% in environment 3; and 15 to 30% in environment 4 (Table 3).Genetic gain, in relation to the overall mean of the experiment, indicates good possibility of gain with selection under these conditions, especially for the environments located in Parana.However, in relation to the controls, lower possibility of genetic gain was observed, when compared with the gain of the overall mean of all the clones of the experiments.
The data presented by Isik et al. (2005) corroborate those reported in this study.According to the authors, the volumes of Pinus taeda clones selected at four years of age in each environment were of 27% and 31% greater than the mean volume of all the clones tested by Isik et al. (2005).Nevertheless, when the authors compared the gain with the families used as control, the former were around 4% to 13%.Pinus taeda breeding programs have increased volumetric yield by 10-30% in relation to the sources not subjected to breeding (Mckeand et al. 2003, Mckeand et al. 2006).
Table 3 shows the small difference between the genotypes used as controls (matrices 161, 162, and 163) and the best clones of the experiments.This indicates that these matrices have good performance in the mean of the environments, and may be considered as plastic and reasonably adapted to the different edaphic and climatic growing conditions.
The comparison of predicted genotypic gains in relation to the commercial control is essential, since the goal of a breeding program is to always improve the mean value of the genetic materials (clones) currently planted for commercial purposes, and not only to improve the mean of the population over time.Therefore, an important challenge is to develop genetic materials and selection criteria that maximize the genetic gain of new materials that surpass the mean value of the commercial control.

Stability and adaptability
The five best clones based on the HMRPGV (Table 4) do not totally coincide with the five best clones according to the order of genotypic values predicted by combined analysis of the environments (Table 3).Coincidence was of 80% among the five best clones, and the order among the coinciding clones was inverted.This lower estimate is associated with selection of the clones that show good performances in both environments, but which are not necessarily the best clones of each environment.The interaction reduces the correlation between the genotypic and phenotypic values, and also reduces the genetic gains with selection (Nunes 2002).This was expected, since the greatest gain is obtained with direct selection for the trait of interest and for the specific environment.The present results corroborate those reported by Martinez et al. (2012) in Pinus taeda families.
When comparing the gains obtained from HMRPGV in relation to the controls (matrices 161, 162, and 163), the mean superiority of these five genotypes was of 33.3% (Table 4).When compared with that predicted in the order of genotypic values of combined analysis of environments (Table 3), also in relation to the control, gain was of 10%.
Individual selection, considering the selection by harmonic mean of the relative performance of genotypic predicted values (HMRPGV), is advantageous for considering the three attributes (productivity, adaptability and stability) (Table 5), taking into account that these new attributes or selection criteria will lead to a more accurate selection (Resende 2007).
Results show that simultaneous selection by adaptability and stability of the genotypic values (HMRPGV) generates 10% additional gain in relation to the control.According to Resende (2007), this occurs because simultaneous selection in the new genetic materials takes advantage of the gain from mean interaction between the environments, which does not occur with the genetic material used as control, since they go through many replications in the trials, and their heritability at mean level tends to be equal to 1.0 in each trial.According to Anputhas et al. (2011), the recommendation PC Dias et al.
of cultivars with broad adaptability and stability is essential for regions with different productive environments, or with distinct climatic seasons.
The selection of the 20 best Picea glauca clones based on height at four years after planting generated mean genetic gain of 4.3% (Wahid et al. 2012), which is lower than that obtained in the present study.According to the authors, this result is considered as important for selection, taking into account that the genetic gain is static and any increase generates gain in selection.Thus, for the present study, selection that considered simultaneously adaptability and stability generated gains close to 10%, and may be used for the recommendation of new clones within the breeding program of the company.Similar result was reported by Sun (2004) when evaluating the adaptability and stability of introduced families of Pinus taeda.
Results show that selection using the volume may be practiced from the fourth year after planting Pinus taeda clones propagated through somatic embryogenesis.Due to the high magnitude of the "g x e" interaction involving Pinus taeda clones propagated by somatic embryogenesis in the two states, a single selection program cannot be adopted, requiring selection of specific clones for the different environments, unless the attributes of adaptability and stability of the clones are used in their selection.Estimated gains confirm the efficiency of the somatic embryogenesis technique in propagation of clones with good yield, aggregating better results to Pinus breeding programs.

Table 1 )
, indicating good ability of PC Dias et al.

Table 1 .
Estimates of genetic parameters for the traits height (Ht), in meters, diameter (DBH), in centimeter, survival rate (sur), and volume (vol), in m 3 in Pinus taeda clones propagated by somatic embryogenesis, at one, three, and four years of age, for the four clonal tests.CVgi (%): coefficient of residual variation; overall mean of the experiment.*Significant by the likelihood ratio test at 5% significance.

Table 2 .
Estimates of genetic parameters for diameter (DBH), height (Ht) and volume (vol) of Pinus taeda clones propagated by somatic embryogenesis, at four ages, in relation to the four clonal tests (two in Santa Catarina and two in Parana) h 2 g: broad-sense herdability of individual plots; c 2 int: coefficient of determination of the effects of genotype x environment interaction; h 2 mg: adjusted heritability to genotypes mean; Acgen: accuracy of selection of genotypes; rgloc: genotypic correlation between performance in the different environments; CVge (%): coefficient of genotypic variation; CVe (%): coefficient of residual variation; overall mean of the experiment.*Significant by the likelihood ratio test at 5% probability.

Table 3 .
Ordering of Pinus taeda clones, propagated by somatic embryogenesis, according to their genotypic values and predicted gains for volume (m 3 ha -1 year -1 ), in combined analysis of environments and in each environment at four years of age g: total genotypic effects; u + g: predicted genotypic value.The highlighted genotypes correspond to the controls.

Table 4 .
Stability and adaptability of genotypic values (HMRPGV) predicted by the BLUP analysis for volume (m 3 ha -1 year -1 ) at four years of age