Selection and genetic gains for juvenile traits in progenies of Hevea

Five yield traits were investigated in three-year-old progenies from open-pollinated rubber trees [Hevea brasiliensis (Willd. ex Adr. de Juss) Muell.-Arg.]. Twenty progenies were evaluated in a randomized, complete block design replicated three times using 10 plants per linear plot at the North Central Experimental Station in Pindorama, Sao Paulo State, Brazil. The characters evaluated included the average yield of rubber, growth vigor, bark thickness, total number of latex vessel rings and latex vessel size. Highly significant (p < 0.01) genetic differences were observed among progenies for most traits. The genotypic variance components accounted for 13.2%, 12.3%, 9.4%, 3.4% and 0.23% of the phenotypic variance for yield, growth vigor, bark thickness, total number of latex vessel rings and latex vessel size, respectively. Heritabilities, as well as genotypic and phenotypic correlations among traits, were estimated. Heritabilities for the above traits at the individual plant level () were 37%, 35%, 69%, 10% and 16%, respectively. Significant positive genotypic and phenotypic correlations were found between the yield of rubber and growth vigor (rg = 0.73, rp = 0.70), bark thickness (rg = 0.70**, rp = 0.75**) and the total number of latex vessel rings (rg = 0.64, rp = 0.80). There was no relationship between yield and latex vessel size, growth vigor or total number of latex vessel rings. Based on these data, selecting the best two out of 20 progenies would result in a genetic gain of 12.3% and 6.8% for yield of rubber and growth vigor, respectively. The two best individual ortets within each progeny would result in a genetic gain of 27.7% and 9.1%, with a total gain of 40% and 16% for these two traits, respectively.


Introduction
The success of a rubber tree [Hevea brasiliensis (Willd.ex Adr. de Juss) Muell.-Arg.]breeding program depends, to a large extent, on knowledge of the genetic aspects involved and on the recognition of traits considered important for plant selection.Information on the genetic variability or heritability of traits that are to be improved is extremely important and of vital importance for prediction of the final outcome.
The rubber tree is a tree crop with a long generative and testing cycle.A single generative cycle requires 4-5 years (excluding 7-10 years for the evaluation of parental performance) and about 20-30 years for one testing cycle from the time of pollination.Thus, one of the most desirable innovations in tree breeding is a technique for predict-ing the breeding value of the trees at the juvenile stage (Varghese, 1999).
The aim of this work was to estimate the genetic variability, heritability coefficients, and the relative genetic gains for the yield traits of rubber tree progenies in order to obtain information necessary for choosing the most appropriate methods of breeding selection.

Materials and Methods
The material for this experiment consisted of seeds from 20 open-pollinated progenies randomly selected from a population base of 100 Asiatic clones.The seeds were sown in the autumn of 1998 in individual polyethylene bags (12 cm x 18 cm x 18 cm) filled with a mixture of 70% humus and 30% sand.The seedlings remained in these pots during the first six months of the growing season and were subsequently planted out at the Pindorama Experimental Station (21°13'S, 48°56'W).The space between rows and between trees within rows was 1.5 m.This region has a humid, tropical climate with a mean yearly temperature of 21 °C and a red yellow podzolic type soil (Lepsch and Valadares, 1976).
The experiments were done using a randomized block design with three replications and 10 plants per plot.A clone trial consisting of a randomized block design with 25 clones, three replications and six plants per plot along with the progenies trial.The dates and procedures were similar to those of the progenies trial.The clone trial was designed to provide an estimate of the environmental variance among plants within plots ( $ σ ew 2 ) (Vencosvsky and Barriga, 1992).

Measurements
The yield of rubber was determined using the principle of the Hamker-Morris-Mann test (HMM) (Tan and Subramanian, 1976), as follows: 30 tappings were made with a half-spiral, alternative daily (1/2S d/3) tapping system 20 cm from the ground.The yield was recorded by cup coagulation.The "coagula" were air-dried for two months until a constant weight was achieved.The yield from each plant was expressed in grams per tapping.
Growth vigor was determined as the girth of the seedling measured once a year.The first measurement at 12 months was of the diameter because the plants were too small to measure growth vigor.Plant diameter was measured 0.50 m from the ground level with a slide clipper.This measurement was converted to growth vigor by assuming that the stem was cylindrical.Another measurement was taken at 24 months at 50 cm above the ground using a measuring tape.
For latex vessel counts, virgin bark samples were removed from the trunk as plugs 20 cm from the ground on the opposite side of the tapping panel.Bark thickness was measured using samples from each seedling.
The total number of latex vessel rings was determined by examining the radial longitudinal sections of the same bark samples.Latex vessel size was expressed as the diameter of latex vessels observed in transverse sections of the same bark samples.Complete details of the procedure are described by Gonçalves et al. (1995).

Analysis of variance (ANOVA)
The analysis using the statistical model below considered all variables (except mean) as random effects.
where y ijk = observed value of the kth plant in the jth replication within the ith progeny, µ = general mean, p i = effect of the ith progeny (p = 1, 2,...20), b j = effect of the jth replication (j = 1, 2, 3), e ij = experimental error associated with the ijth plot and d ijk = effect of the kth plant within the ijth plot.
Estimates of the components of genotypic and phenotypic variances were obtained by the mathematical expected mean squares from the analyses of variance (Table 1), according to Vencovsky and Barriga (1992).Thus, the phenotypic variance due to differences among plants within plots was: , which can be separated into genetic ( $ ) σ gw 2 and environmental ( $ σ ew 2 ) variances among plants within plots.
Partitioning the component $ σ w 2 into $ σ gw 2 and $ σ ew 2 suggested that it would be possible to use information from the mean square within plots of the clones trial (Table 2) to provide estimates of the environmental variance within plots ( $ ) σ ew 2 .Thus, the environmental and genetic variances among plants within plots were obtained by the mean squares (Tables 3 and 4).
Estimates of the environmental variance among plots ( $ σ e 2 ) and of genotypic variance due to differences among 208 Gonçalves et al.
Table 1 -Degrees of freedom (df), mean square (MS) and expected mean square (EMS) for estimating the variance due to differences among plants within plots in a clonal trial of rubber trees.

Sources of variation
variance due to differences among plants within plots, $ σ e 2 = environmental variance among plots, σ eg = genotypic variance due to differences among progenies, r = number of replications, f = number of progenies, n = number of plants per plot (harmonic mean).
progenies ( $ σ g 2 ) were obtained through the expected mean square, as shown in Table 1.
The total genotypic variance ( $ σ G 2 ), phenotypic variance based on progeny means ( $ σ p 2 ), and phenotypic variance based on individual plants ( $ σ p 2 ) were obtained using the expressions:

Heritabilities and gain estimates
The heritability coefficients at the individual plant level ( $ h i 2 ), for selection within progeny ( $ h w 2 ), and for the among progenies mean ( $ h g 2 ), were estimated using the expressions (Vencovsky and Barriga, 1992): Genetic gains for juvenile traits in progenies of Hevea 209 Table 2 -Degrees of freedom (df), mean square (MS) and expected mean square (EMS) for estimating the variance due to differences among plants within plots in a clone trial of rubber trees.The coefficients of genotypic variation (CV g $ %) were estimated using the formula described by Vencovsky (1983):

Sources of variation
Genetic gain from selection among progenies (G g ) and within progenies (G w ) was also computed using Falconer and Mackay's (1996) formula: where k = the selection differential in a standard measure, $ σ P = the phenotypic standard deviation of the progeny means, and $ σ w = the phenotypic standard deviation among plants within plots.
The genetic gain expressed as a percent of the mean obtained from among and within progenies was estimated using the formula: where x is the general mean.

Correlations coefficients
To examine the relationship among the yield traits of rubber trees, genotypic (r g(xy) ) and phenotypic (r p(xy) ) correlation coefficients were calculated according to Kempthone (1966), as follows: The significance of these correlations was tested using the table in Fisher and Yates (1971) for simple correlations for 5% and 1% probabilities.The genotypic and phenotypic correlation coefficients were tested using the degrees of freedom of the error.

Genotypic variation
Except for the total number of latex vessel rings, highly significant among-progeny differences were de-tected by the F test (Table 3).This variability is an essential condition for establishing a genetic breeding program and could be effectively exploited to increase rubber tree production.
The genotypic variation coefficient, which expresses the amount of existing genetic variation as a percentage of the general mean, was higher for the yield of rubber (22%) than for growth vigor (10.2%), bark thickness (8.7%), latex vessel size (1.4%) and total number of vessel rings (0.96%).These results confirmed the F test results for progeny differences and characterized the yield of rubber as a more suitable trait expressing the genetic variability in this population.These results also agreed with those reported by Gonçalves et al. (1998) and Boock et al. (1995) for the same traits.
Table 4 shows the clone trial analysis of variance.Except for latex vessel size, significant among-clone differences were detected by the F test for all traits.These results reinforce the variability observed in the progeny trial, an essential condition for establishing a genetic breeding program.The main purpose of the clone analysis of variance was to supply an estimate of the environmental variance among plants within plots.

Progeny variance components
The genotypic variance components among progenies accounted for 13.2%, 12.2%, 9.4%, 3.4% and 0.83% of the phenotypic variance based on individual plants for bark thickness, growth vigor, yield of rubber, latex vessel size and latex vessel rings, respectively (Table 5).The contribution of genotypic variance among progenies accounted for $ σ 2 g = 13.7% for bark thickness of the total genotypic variance and was substantially smaller than that for growth vigor ( $ σ 2 g = 35.4%).In addition, the estimates of the growth vigor were substantially greater than for the other traits.

Heritabilities
The mean heritability among progenies for the yield of rubber and growth vigor at the individual plant level (h i 2 ) were relatively low, with values of 0.37 and 0.34, respectively (Table 6).These data agreed with the results of Costa et al. (2000a,b), Gonçalves et al. (1998Gonçalves et al. ( , 1999)), Boock et al. (1995) and Moreti et al. (1994).Gonçalves et al. (1998Gonçalves et al. ( , 1999) ) reported individual plant heritabilities of h i 2 = 0.37 and h i 2 = 0.51 for the yield of rubber, whereas Costa et al. (2000a) and Boock et al. (1995) reported values of h i 2 = 0.24 and h i 2 = 0.35, respectively, at Votuporanga.In addition, Costa et al. (2000b) obtained a value of h i 2 = 0.11 at Jaú.For growth vigor, values of h i 2 = 0.48 (Gonçalves et al., 1999) and h i 2 = 0.31 (Moreti et al., 1994) were obtained at Pindorama and h i 2 = 0.47 (Boock et al., 1995) at Votuporanga.These results indicate that these two traits have moderately low heritability estimates and imply that there is a need to increase the genetic variability for yield and growth vigor in the population studied here.
Values of h g 2 = 0.89 (Costa et al., 2000a) and h g 2 = 0.81 (Gonçalves et al., 1999) were reported for growth vigor at Votuporanga and Pindorama, respectively.These highly significant heritability estimates for progeny means (h g 2 ), together with the extensive variation seen within the rubber tree population, suggest that an effort to increase yield, growth vigor, total number of latex vessel rings and latex vessel size through progeny selection would be rewarding, as also indicated by Gonçalves et al. (1999).
The mean heritability among progenies for other traits, such as the total number of vessel rings and latex vessel size, were low, i.e., 10% and 16% at the individual plant level (h i 2 ), 10% and 13% within progenies (h w 2 ) and 13% and 42% among the progeny means (h g 2 ), respectively.

Expected genetic gains
The expected genetic gains for the traits studied at different levels of selection intensity are shown in Table 7.If the best six progenies were selected and planted at sites similar to that used in the experimental trial, a genetic gain among progenies for yield and growth vigor of 8.3% and 4.6%, respectively, could be achieved.When only the two best progenies (i.e., two out of 20 progenies) were selected for yield and growth vigor with a selection intensity of 10%, the gain in production would be 29.1 g and 188.6 mm, which is 12.3% and 6.8% greater than the overall progeny means.Even smaller increases than these would result in appreciable improvement if the plantings were large enough.
On the other hand, if the two best individuals were selected within progenies and then asexually multiplied and planted at the same site, the genetic gain for yield and growth vigor would be about 27.7% and 9.1%, respectively.For yield, growth vigor and bark thickness, the gains would be 27.7%, 9.1% and 18.4%, respectively.Considering combined selection among and within progenies, the total genetic gains would be 40%, 16% and 29.5% for yield, growth vigor and bark thickness, respectively.

Correlations
Genotypic and phenotypic correlations between yield and other characters, such as growth vigor and bark thickness, were generally significant (Table 8).Progenies with high yield and high growth vigor often showed a highly significant positive correlation with bark thickness.
The influence of bark thickness on yield and growth vigor was not immediately evident in the field.High growth Genetic gains for juvenile traits in progenies of Hevea 211  2 Expressed as the girth of the progenies measured 20 cm from the ground.
Table 7 -Expected gains from different levels of selection and percentage of the mean, among (G g ) and within (G w ) progenies, and selection total (G g + w ) for five traits in 20 open-pollinated rubber-tree progenies.

Traits
Selection level 1 From selection In percent of the mean  vigor and increased bark thickness correlated with yield to varying degrees.A positive genotypic and phenotypic association between the average yield and the number of latex vessel rings was not evident.The lack of a significant genotypic and phenotypic correlation between growth vigor and the total number of latex vessel rings and between yield and latex vessel size indicated that there would be a low genetic gain for all of these traits, even if selection were undertaken for only one trait.

Breeding strategy
The results of this study provide a basis for developing a program to improve rubber tree selection.The "progeny plus within-progeny selection" method is currently the predominant form of selection used in most advanced tree improvement programs (Zobel and Talbert, 1984).This method consists of selecting the best progenies and the best individuals among these progenies.We intend to adopt this method to improve the yield of rubber, growth vigor and total number of latex vessel rings.Since there is significant genetic correlation among the three characters, direct selection can be used in this case.For this purpose, "massal selection" will be used, i.e., growth vigor will be selected first.When a desired level of growth vigor has been obtained, breeding efforts will be concentrated on yield.However, if a simultaneous improvement in yield and growth vigor were desired, then an independent culling method would be recommended.
due to differences among plants within plots, r = number of replications, c = number of clones, n = number of plants per plot.

where
Cov g(xy)  and Cov p(xy) = the genotypic and phenotypic covariances for traits x and y respectively, σ gx 2 and σ px 2 = genotypic and phenotypic variances for trait x, and σ gy 2 and σ py 2 = genotypic and phenotypic variances for trait y.

Table 3
-Degrees of freedom (df) and mean squares from ANOVA, general mean, and experimental (CV e %) and genotypic (CV g %) coefficients of variation for five traits in 20 open pollinated rubber tree progenies.*p < 0.05, **p < 0.01, n.s.= not significant. 1g/t/t -grams per tree per tapping. 2 Expressed as the girth of the progenies measured 20 cm from the ground.

Table 4 -
Degrees of freedom (df) and mean squares from ANOVA, general mean, and experimental (CV g %) and genotypic (CV g %) coefficients of variation for five traits in 25 rubber tree clones.

Table 5 -
Estimates of genotypic and phenotypic components of variance for five traits in 20 open-pollinated rubber tree progenies.Expressed as the girth of the progenies measured 20 cm from the ground. 2

Table 6 -
Heritability coefficients for five traits in 20 open-pollinated rubber tree progenies.

Table 8 -
Estimates of genotypic (r g ) and phenotypic (r p ) correlation coefficients based on individual plants among five in 20 open-pollinated rubber tree progenies.
*p < 0.05, **p < 0.01. 1 Expressed as the girth of the progenies measured 20 cm from the ground.