GENETIC VARIATION AND REALIZED GENETIC GAIN FROM RUBBER TREE IMPROVEMENT

Breeding cycle in rubber extends to 20–30 years between pollination and yield assessment, distributed over three selection stages. Five hectares of small scale trial of rubber tree [Hevea brasiliensis (Willd. ex Adr. de Juss.) Muell.-Arg.], was established in the Northwestern region of São Paulo State, Brazil. The population comprises 98 clones mostly derived from intensively selected plus tree in the natural forest of rubber tree in the Amazon. Three open pollinated progeny tests were established in three experimental stations. Seedlings from 98 progenies including a commercial check (CC) were planted in each one of the three locations. Assessments were made when the plants were one, two and three years old. The variation among progenies for girth was highly significant in all locations examined. In the combined analysis of variance over three locations, differences among progenies were also detected, while progeny × location interaction effect was not significant. Narrow sense heritability estimates on individual tree basis (hi ) were variable depending on the characteristic, age of assessment and experimental location. Realized genetic gains were calculated for the plant characteristics at the age of three years, by comparing the performance of improved (selected) materials to unimproved materials (CC). The total genetic gain from the genetically tested first generation clone population at Votuporanga is estimated as 25% for girth, 14% for rubber yield and 25% for bark thickness. Improvement of rubber tree by selection, establishment of clonal population (isolated garden) and progeny testing is a very promising and profitable operation.


INTRODUCTION
Rubber tree [Hevea brasiliensis (Willd.ex Adr. de Juss.)Muell.-Arg.] is native to rain forests of the Tropical region of the Great Amazon basin of South America.This area, between Equator and 15º S is distinctly characterized by its flat land with altitudes not exceeding 200 m with a wet equatorial climate (Strahler, 1969).Monthly temperature is 25º to 28ºC with abundant rainfall of more than 2000 mm/year, without any real dry season.The amount of precipitation is at least twice the evaporation losses on yearly basis (Pushparajah, 2001).
Hevea the prime source of commercial rubber, is a deciduous perennial tree of the family Euphorbiaceae.Rubber is the strategic raw material for more than 40,000 products, including 400 medical devices (Mooibroek & Cornish, 2000).Primarily due to its molecular structure and high molecular weight (> 1 million daltons) it has resilience, elasticity, abrasion resistance and impact resistance that cannot easily be mimicked by artificially produced polymers.
Breeding cycle in rubber extends to 20-30 years between pollination and yield assessment, distributed over three selection stages.This justifies standardization of early selection methods to optimize and shorten the cycle as much as possible.One component of early selection is the identification of traits at young age that have correlated responses with yield at maturity, and the other is combined management of different selection stages to improve the accuracy of estimation of genetic value.
Breeding methodologies utilized for maximizing genetic gain are based on breeding objectives with the specific aim of providing farmers with high yielding clones.Such methodologies are backed by the theory of quantitative genetics, which derives clones well adapted to a given environment.Elements of breeding methodologies are available with major contribution from Tan (1987); Simmonds (1989); Gonçalves et al. (1998); Gonçalves et al. (2005) and Priyadarshan & Clément-Demange (2004).
The objective of this investigation is to study the genetic variability for growth and quality characteristics in Hevea to identify clones with high breeding value for selection and to estimate realized genetic gain from first stage clonal population.

Plant material
Open pollinated seeds were collected from all 98 phenotypically selected clones most of them derived from intensively selected plus tree in the natural rub-ber tree forest in the Amazon region.All seed from a clone, regardless of the ramets from which they were collected were considered to be members of the same open pollinated progeny, i.e., they all shared the same genetic mother.Seeds were sown in polyethylene bags.A year later three experimental plantings were established.

Experimental locations
At each location, 98 progenies plus a control, also called commercial check (CC), GT 1 progenies were planted in randomized complete block design including three (Pindorama and Jaú) or four (Votuporanga), replications.Progenies plus CC were randomly assigned within replications in 10 plant row plot arrangement for Jaú and Pindorama and 8 plant row plots for Votuporanga.The rubber tree cultivation regions in São Paulo State are characterized as: (i) Jaú experimental station (22º17' S, 48º34' W, altitude 580 m).The soil is a medium-textured, deep and well drained Rhodic Hapludox, with a flat topography.An Aw (köppen) climate predominates in this region, with a defined dry season, annual mean temperature of 21.6ºC, average humidity of around 70% with extremes of 77% in February and 59% in August.The mean annual rainfall is around 1,344 mm.
(ii) Pindorama experimental station: latitude 21º13' S, longitude 48º56' W, altitude 560 m, with a medium textured, deep, abruptic and well drained Typic kandiustalf.The climate is tropical, with a wet summer and a dry winter period.Mean monthly temperatures vary from 16.6 to 28.9ºC, and annual mean rainfall is 1,390 mm.The period from October to April usually has favorable precipitation for growth and production.Precipitation deficiencies and low temperatures occur from May to September.
(iii) Votuporanga experimental station: latitude 20º25'S, longitude 49º50' W, altitude 450 m.The soil is a medium textured Typic Kandiudalf, and the climate, tropical, with a hot wet summer and dry winter, with low temperatures and rainfall.Mean monthly temperatures varies from 20 to 25ºC and annual mean rainfall ranges from 1,087 to 1,537 mm.The winter drought varies from four to six dry months, with an average water deficiency of 180 mm.
Since the geomorphology of the experimental planting at Votuporanga was quite heterogeneous, to reduce the variability within replications, the number of replications was increased with equivalent reduction in plot size.The spacing in all plantings was 2 × 2 meters.The characteristics of the locations and the experimental plantings are shown in Table 1.

Measurements
Girth was measured in all three locations at the ages of one, two and three years.Rubber yield and bark thickness was measured only at the age of three years.In the first year, diameter was recorded because plants were too small to measure the girth.Measurements were converted into girth assuming that plants were measured 0.20 m above ground level by a slide caliper.The other two girth measurements were recorded 0.50 m above the ground.
Yield determination as dry rubber was carried out using the Hamker Morris-Mann test (HMM) (Tan & Subramanian, 1976), as follows: thirty tappings were done with a half-spiral alternate daily (1/2S d/2) tapping system at a height of 20 cm from the ground.The yield was recorded by cup coagulation.The "coagula" were air dried for two months until they reached a constant weight.The results were expressed in grams per tapping.Virgin bark samples were removed from the trunk as plugs at 20 cm high on the opposite side of the tapping panel.Bark thickness was measured using the sample for each progeny.

Statistical evaluation
Data were checked for suspected outliers (measurements outside the normal values) before analyses.Such values may arise from various reasons, as recording errors during measurements or mechanical damages of plants in the growing locations.Analysis of variance for each character measured or assessed were conducted in two stages due to missing plants within plots and the need for the estimation of the within plot variance, following well established procedures (Namkoong, 1979;Becker, 1984).Initially the mean squares within plot variance (σ 2 w ) were estimated from individual plant data.Then, analyses based on plot means were made applying the general model of two way classification: where : Y ij = the mean plot value of the j th progeny that is in the i th replication all Y ij values are treated as if they had the same number of observations; µ = overall mean; r i = effect of i th replication; f j = effect of j th progeny; rf (ij) = interaction effect of progeny × replications; w ij = the mean of the n ij deviations; The variance of plot means , where h n is the harmonic mean of the number of plants within plots.The within plot mean square variance (σ 2 w ) was used for estimating heritability values, while the average variance of plot mean (σ 2 w /n h ) is used in the analyses of variance for testing the replications × families interaction effect.

Combined analyses of variance over locations
Genotype × environment interaction arises from the fact that some genes may have effects that differ from one environment to another.Such interactions, if exist, may cause decreased performance when a population selected under one environment is Sci. Agric.(Piracicaba, Braz.), v.66, n.1, p.44-51, January/February 2009 used under another.The size and extent of the interaction is very important for the movement of recombinant material of the rubber tree species.Therefore, data girth measured at the ages one, two and three years old, common to all (three) experiments, were pooled over locations and combined analyses of variance were performed following the procedures outlined in detail by Cochran & Cox (1957).Since the numbers of replications were not equal in all locations (three in Pindorama and Jaú and four in Votuporanga) and the number of trees within plots also varied, the analyses for getting the effect of locations, progenies and progenies × locations were based on plot means.
Genotype × environment interaction (G × E) effect at the progeny level was also tested for significance using the type B (r b ) genetic correlation (Hodge & White, 1992;Sierra-Lucero et al., 2002).This correlation express the genetic relationship that may exist between measurements of the same character taken in different environments and it is expressed as: ); Values of r b near one indicates no (G × E) interaction, while values near zero indicates highly significant interaction.The variance components of progenies (σ 2 f ) and progeny × location interaction (σ 2 fp ) were estimated from the pooled over location ANOVA.

Genetic parameter estimates
Mean squares were equated to expected mean squares and variance components and heritability estimates were obtained.The degree to which progeny members are closely related is proportional to the degree to which progenies differ (Namkoong, 1979).Thus the variances among progenies (σ 2 f ) obtained from the analyses are interpreted as one quarter of additive genetic variance.
Narrow sense heritability estimates on individual plant basis (h 2 ) for each one characteristic and location were estimated as the ratio of the additive genetic variance (σ 2 A ) to total phenotypic variance (σ 2 ph ): The values obtained from the above formula are applicable in the specific environmental conditions under which they have been estimated and are biased upwards when applied to different environmental conditions.

Estimation of realized gain
In clonal population the clones originally selected in the first phase are crossed and the progenies produced are evaluated in progeny tests and estimates of the genetic value of the parental clones are made.The clones, which prove to be poor parents are culled from the clonal population and only the best (genetically tested), are used for future breeding.Realized genetic gains were estimated for both stages of the selection program i.e., for the unrogued and rogued clonal population and selected clonal population of Votuporanga.Since unimproved commercial checks (CC) were included in all plantings, the gain for the first stage of selection was computed by subtracting the mean of the unimproved (CC) from the overall mean of the progenies of the selected clones included in the clonal population; the difference was expressed in percentage of the commercial check.After culling the clonal population on the basis of progeny testing (genetic thinning) an additional gain is obtained.This gain was calculated by subtracting the overall progeny mean of the test (excluding CC) from the mean of the reselected progenies, to be maintained in the clonal population.Finally the total realized genetic gain from rogued clonal population was taken by the addition of the gain of the two stages of selection (mass selection plus progeny testing).

Growth characteristics
Survivals at the stage of three years were variable among progenies and locations.At Pindorama the survivals among progenies ranged from 81% up to 100% with overall mean of the test planting 96%.Twelve progenies had 100% survival at the age of three years.This high survival is related to favorable growth conditions and specifically to very high moisture holding capacity of the soil.In Votuporanga with less favorable conditions the survivals varied among progenies from 55% up to 100% with overall mean 85%.Finally in Jaú, survival was worst ranking among families from 50% up to 90% with overall mean 70% (Table 2).Analyses of variance (not presented) have shown that the differences among progenies within locations were significant (p < 0.05), while the differences among locations were highly significant (p < 0.01).Total girth at the age of three years was related to survival rate.It was higher (25.90 cm) at Votuporanga with the highest survival and lowest (17.74 cm) at Jaú experimental planting.The differences among progenies for one, two and three years old and rubber yield at the age of three years old, were statistically significant.Large variability among open pollinated progenies of Hevea for growth traits, at the age of three years old, has been also reported by Costa et al. (2000a).

Variance components and heritability estimates
Knowledge of variation and the degree of inheritance of the characteristics of interest are the fun-Sci.Agric.(Piracicaba, Braz.), v.66, n.1, p.44-51, January/February 2009 damental basis upon which a well designed Hevea improvement program must be based.Heritability estimates were variable depending on the location, age and characteristics to which they refer (Table 3).
In Pindorama the h i 2 for girth were 0.50, 0.62 and 0.67 at the ages of one, two and three years old, respectively.The h i 2 for rubber yield at the age of three year was found 0.64 and that for bark thickness 0.98.It seems that in this experimental planting the h i 2 values increase as plants become older.This was expected since the environmental variation is reduced, as plants become older.Similar results have also been reported in Hevea (Costa et al., 2000b).They found that girth heritability increased from 0.60 at the age of three years to 0.80 at the age of five years.
In the planting of Jaú the h i 2 values for girth were 0.69, 0.79 and 0.67 at the ages one, two and three years old, respectively, while the corresponding h i 2 values for three years old rubber yield was 0.31, and that for bark thickness 0.68.Heritability estimates of girth at the Votuporanga experimental planting were very close to those obtained from Pindorama (

3). Heritability values (h i
2 ) for growth characteristics in rubber tree have also been studied by Gonçalves et al. (2006) at the age of seven years; from the analysis of 67 different genotypes including five locations, they estimated h i 2 values of 0.65, 0.86 and 0.87 for girth, rubber yield and bark thickness, respectively.These values are in close agreement with those found in the present study.The higher h i 2 values for girth at the ages of 1 (0.69) and 2 (0.79) years, estimated in the Jaú planting, may be the result of the higher location quality of this location in comparison with the other two or the better location uniformity within replications.
Early natural selection against the slower growing plants is expected to be stronger on better sites with the consequence of the reduction of the environmental variance.Gonçalves et al. (2005) working on mature trees found positive correlation (r = 0.54) between h i 2 and mean annual increment (r = 0.67), indicating that location productivity influences positively h i 2 values.However, there are contradictory results found in the literature.Costa et al. (2000a) found no relationship between h i 2 values and location index in rubber and concluded that the degree of genetic control and location uniformity were the same on location of both high and low girth.Narrow sense heritability for girth estimated from the pooled analyses of variance across the three locations was 0.79, 0.82 and 0.89 at the ages one, two and three years old, respectively (Table 4).The differences among progenies and among locations were highly significant, while the progeny × location interaction effect was not significant in all ages.This indicates that although there are some differences in the rankings of the progenies, the faster growing (on the average) remain in the top and the slower are found in the bottom.
The type B genetic correlation (r b ) at the progeny level was found close to unity, indicating that progeny ranks are relatively stable across locations.This is in agreement with not significant progeny × location interaction effect found from the pooled analyses of variance.Genotype × environment interaction is not a significant source of variation and therefore the clone selected from the clonal population of Votuporanga can be freely used over the environmental conditions of the three locations that are representative of the broad area of the São Paulo State plateau.Although there are strong interactions for a few progenies the genetic gain does not seem to be influenced when the best progenies are selected on the basis of average performance at all locations.Although the results obtained are based on relatively young progeny tests (three years), they can be used for preliminary decisions, such as initial selection of the clonal population.To increase the efficiency of the breeding and testing Hevea program at Instituto Agronômico (IAC), a research institute in São Paulo State, Brazil, Gonçalves et al. (2005) analyzed 12 first generation progeny tests to find the optimum time for growth assessment.Under reasonable assumptions for age-age correlation and heritability changes over time, they concluded that the expected genetic gain per year in the breeding program was the greatest for selection between, two and three years.Since gain rates will be heavily influenced by time, early selection for girth and for recombinant development will be advantageous (Namkoong, 1979).However, these results must be interpreted with caution since age-age correlation is not always precisely estimated and the performance of the genotypes at young ages, in most of the cases, is imperfectly related to that at maturity (Olson & Lindgren, 2001).It is of interest to keep in mind that juvenile selection, is a form of indirect selection, the efficiency of which is determined by heritabilities at the juvenile and mature age and the correlation that exists between the characteristics measured at these two ages.

Realized genetic gain
Realized genetic gain was estimated at the age of three years.Data for girth, rubber yield and bark thickness common to three locations was pooled and progeny means were obtained over the three locations.The overall progeny mean values were 11.38 cm, 0.84 cm and 3.80 mm for girth, rubber yield and bark thickness, respectively.The mean values for the commercial check (control) were 9.16 cm girth, 0.53 g for rubber yield and 2.12 mm for bark thickness.Therefore the realized gain for the selection made in clone population is estimated to be 14.36% for girth, 8.20% for rubber yield and 15.10% for bark thickness (Table 5).
Economic studies (Bergman, 1968) have shown that 2 to 5% increase in yield alone over commercial planting is enough to justify the cost of a rubber tree improvement program.Therefore the realized gain (10%) for girth, in the present study, resulting from removing makes it clear that rouging is a beneficial part of a tree improvement program and must be practiced if optimal expectations from the first generation clonal population is to be obtained.The total gain from the first generation genetically thinned (roguing) clonal population is summed to 10% for girth, 6% for rubber yield and 10% for bark thickness (Table 5).Since the intensity of the second stage of selection was quite low (80% of the clones were saved and only the rest 20% were removed), higher gains should be expected if this intensity increases.However there is a need to keep broad the genetic base in the rogued clonal population, in order to provide adequate genetic diversity in the improved commercial plantations.This is especially needed in Hevea, with very long cycle of yield (over 40 years), that require well adapted trees (clones) with adequate genetic variation, capable to respond successfully to environmental conditions.Reduction of the genetic base of the seed produced can also be a result from differential fertility of the clones and absence of synchronous flowering.
Although the results of the present study are not directly comparable with the available literature because of different species, age of evaluation, charac-teristics of interest and environmental conditions, some important results found in the literature are discussed here.Estimates of realized genetic gain in rubber tree have been reported in the past by many investigators (Moreti et al., 1994;Boock et al., 1995;Gonçalves et al., 1995Gonçalves et al., , 1996Gonçalves et al., , 1998)).Costa et al. (2000b) have reported results from an extensive study including three open pollinated progeny tests of the rubber tree program of the IAC in São Paulo State.At the age of three years, realized genetic gain for rubber yield varied within and among locations in the different selection methods.When the accuracy values associated to genetic gain are greater, the expected progress with selection was also greater, i.e. the greater the precision in selection the greater the gain.
The realized genetic gain in the present study has been estimated from relatively young materials (three years).Since the performance of genotypes at young ages in many cases are imperfectly related to that at maturity, further evaluation will be made as the trees become older.

Table 1 -
Details of experimental locations and planting dates from three different test locations where 98 Hevea progenies were evaluated in São Paulo State, Brazil.

Table 2 -
Overall mean values for girth at the age of one, two, and three years, rubber yield, bark thickness and survival of three open experimental plantings where 98 Hevea progenies were evaluated in São Paulo State, Brazil.
1 Values in parenthesis are progeny means ranges.

Table 3 -
Estimates for girth at the age of one, two and three years, rubber yield and bark thickness variance components and narrow sense heritability (h i 2 ) on individual plant basis of three open pollinated experimental plantings, where 98 Hevea progenies were evaluated in São Paulo State, Brazil.

Table 4 -
Estimates of variance components and narrow sense heritability (h 2 i ) for girth at ages one, two and three years rubber yield and bark thickness from the combined analyses of variance of three open pollinated experimental plantings where 98 Hevea progenies were evaluated in São Paulo State, Brazil.1 σ 2 g , σ 2 gl , σ 2 w are estimates of variance components for progenies, progenies × locations and within plot, respectively. 2p < 0.01.

Table 5 -
Realized genetic gain for girth, rubber yield and bark thickness from two stages of selection (mass selection plus progeny testing) at the age of three years (Pindorama, Jaú and Votuporanga tests combined).