QUANTITATIVE VARIATION IN THE TROPICAL MAIZE POPULATION , ESALQ-PB 1

Good yield and intermediate plant height, ear height, and tassel size characterize the maize population ESALQ-PB1 as an outstanding breeding population. Estimates of genetic parameters are reported for 13 traits: plant height (PH), ear height (EH), ear placement (EP), tassel length (TL), tassel weight (TW), tassel branch number (TB), ear weight (EW), total grain weight (GW), ear length (EL), ear diameter (ED), kernel row number (RN), kernel number per row (KR) and prolificacy (PR). Results refer to one location and one year. Genetic variation was detected for all traits, and the estimates of the additive genetic variance are presented. The coefficients of heritability (individual basis) varied from 0.14 to 0.72 and were considered high for PH, EH and TB; intermediate for EP, TL, TW, EL, EP, ED and RN, and low for EW, GW, KR and PR. The coefficient of heritability (progeny mean basis) showed approximately the same trend and varied from 0.40 to 0.75. The highest expected gain from selection was for TB (27% per cycle) under mass selection and for TW (16.4%) under progeny selection; the lowest expected gain was for ED either for mass selection (1.9%) or progeny selection (2.9%). Additive correlation coefficients (r A ) of 0.5<r A <0.6 were found for PH or EH and yield traits (EW and GW) and its components EL and KR, and 0.10<r A <0.44 for PH or EH and tassel traits. For the tassel traits the highest correlation was r A = 0.63 (TB.TW). EP was positively correlated with TB, TW, PH, EH and PR; and negatively correlated with ED and RN. The expected correlated responses in several traits after selection for GW, EH and TB are given.


INTRODUCTION
Maize is a typically allogamous species with prevalence of cross pollination.The amount of self-fertilization in an ordinary field of maize is negligible (Kiesselbach, 1922).Therefore, a maize population is seen as a group of individuals that cross randomly among themselves in a panmitic way so that male and Sci.Agric.(Piracicaba, Braz.), v.65, n.2, p.174-182, March/April 2008 female gametes unite at random.The probability of an individual in the next generation be homozygous by descent is 1/2N (N is the population size) and a population sufficiently large can be maintained in equilibrium if no other forces (mutation, gamete migration, selection, random drift) lead to departure from its basic structure.
The genetic structure of a maize population under panmixy is very complex (Shull, 1908).For quantitative traits that are polygenically inherited the complexity arises from the great number of genotypes within a population and from allelic, non-allelic and genotype by environment interactions.For breeding purposes, the genetic structure of maize populations for quantitative traits has been studied on the basis of statistical parameters (e.g., mean, variance, correlation coefficient, and others).The basic concepts and statistical methods for estimation of genetic parameters are given by Cockerham (1963), Hallauer & Miranda Filho (1995), Falconer & Mackay (1996), Lynch & Walsh (1997) and Cruz & Carneiro (2003).
Parameter estimates refer to a particular population of genotypes (reference population) for a defined population of environmental effects (Cockerham, 1963); the estimates can, however, be compared among populations for their relative magnitudes and be used as a guide for effective handling in breeding programs.Parameter estimates for some important traits in maize populations were summarized in several opportunities (Miranda Filho, 1985;Vencovsky et al., 1987;Hallauer & Miranda Filho, 1995).More recently, genetic parameters also estimate for others specific populations by Paterniani et al. (2004), Carvalho et al. (2005), Yazdi-Samadi et al. (2004), El-Bagoury et al. (2005), Zivanovic et al. (2005) and Salami et al. (2007).
The purpose of this study was to report about parameter estimates of 13 quantitative traits in the maize population ESALQ-PB1 and its potential for breeding.Particular attention was given to tassel branch number and ear height because ESALQ-PB1 was used as a base population for divergent selection of these traits.

MATERIAL AND METHODS
The reference population, ESALQ-PB1, is a composite formed from crosses among seven short plant varieties (Miranda Filho, 1974).After three generations of recombination and homogenization, it has undergone one cycle of stratified mass selection and one cycle of half-sib family selection.One thousand half-sib families were evaluated in the last cycle and high selection intensity (2%) was used.After recombination of the selected 20 half-sib families, 147 open-pollinated ears (half-sibs) were used to represent the base population of this study.
The 147 half-sib families were evaluated in three 7 × 7 lattice experiments with four replications, at Piracicaba-Brazil (22°39' S and 47°35' W, and 597 m of altitude), in 1978/79.Plots were 5 m long, spaced 1 m apart, with 25 plants after thinning.The following traits were taken for analysis: PH -plant height (m); EH -ear height (m); EP -ear placement; TL -tassel length (cm); TW -tassel weight (g); TB -tassel branch number; EW -ear weight (g); GW -grain weight (g); EL -ear length (cm); ED -ear diameter (cm); RN -kernel row number; KN -kernel number per row; and PR -prolificacy.The traits EL, ED, RN, KN and PR were taken in only two experiments (98 families).Actually, this is a selection project and the experimental layout was designed according to this purpose.The quantitative traits analyzed herein took only part of the experiments, so that a sample of size five was taken within plots to measure plant, tassel or ear characteristics.
Because the low efficiency of the lattice design in relation to the randomized complete block design, all the analyses of variance and covariance were performed as complete block design with plot means and the variation within plots was included.The basic model is: where Y ijk is the observation of the k th plant of the i th progeny in the j th replication; m the overall mean; p i the random progeny effect (mean: 0; variance: σ² p ); b j the random block effect; e ij the experimental error (mean: 0; variance: σ² e ), and s ijk the effect of the k th plant (mean: 0; variance σ² w ) in the ij th plot.The analyses of variance and covariance are shown in Table 1 with the mean squares (M 1 and M 2 ), mean products (P 1 and P 2 ) and their respective expectations in terms of variance and covariance components.The following estimates were obtained of the individual analyses grouping: • Genetic variance among progenies..The estimates involving PH, EH, EP, TL, TW, TB, EW and GW, were calculated on basis in 147 families (three experiments).For EL, ED, RN, KN and PR such as its correlations with PH, EH, EP, TL, TW, TB, EW and GW, were considered only 98 families (two experiments).

RESULTS AND DISCUSSION
Means of half-sib families and their respective ranges for 13 traits in the maize population ESALQ-PB1 are shown in Table 2.For plant height (PH) and ear height (EH) the population should be classified as intermediate, as compared to other brazilian populations.Ear placement (EP) indicates that on the average the ear is located in the upper half of the stalk, but the range of variation (0.475 to 0.535) indicates that EP can be reduced through selection.The tassel characteristics did not differ greatly from other populations (Geraldi et al., 1985;Souza Jr. et al., 1985).Tassels had an average length (TL) of 40.9 cm and the average weight (TW) was 5.9 g.The tassel branch number (TB) deserves special attention because it is the primary trait used in the selection for tassel size   (Mock & Pearce, 1975).Tassel branch number averaged 18.4 branches per tassel, varying from 12.6 to 23.8, which would be considered as intermediate values as compared to other populations.Average of 26.5 branches per tassel was reported for 338 samples representing races of maize from several countries, varying from 17.7 (Bolivian races) to 42 (Colombian races) (Paterniani, 1981).Averages of 18.1, 22.4 and 18.7 branches/tassel in three broad-base populations (Centralmex, Flint Composite and Dent Composite, respectively) also were reported (Geraldi et al., 1985).Variation from 17.5 to 33.2 for TB are found in the maize population SUWAN (Souza Jr. et al., 1985).Average of 15.8 branches/tassel, varying from 6.9 to 22.5, were found among 15 exotic populations (Miranda Filho, 1992), while the mean for the population ESALQ-PB1 was 20.4.
The means for EW (185.5 g) and GW (150.8 g) per plant were greater than the average of 107 mean estimates of Brazilian populations (Miranda Filho, 1985) and the population RPM (Resistant to Phaeosphaeria maydis) and RET (Resistant to Exserohilum turcicum) (Miranda Filho & Gorgulho, 2003) and Composite I-Mo (Paterniani et al., 2004).The greater range of variation (Table 2) for these traits is an expression of their high genetic variability, however other populations of broad genetic basis present larger estimates (Cárdenas, 2005;El-Bagoury et al., 2005).The EL varied from 13.9 to 19.8 cm, averag-ing 17.5 cm.For ED a small variation (4.5 to 5.1 cm) was observed, while the RN and KR showed a wide variability, ranging from 12.6 to 16.8 and 26.4 to 41.4, respectively.Prolificacy, as measured by the average number of ears per plant under a population density of 50,000 plants ha -1 , varied from 1.0 to 1.4 among half sib families.
The experimental coefficient of variation (CVe) varied from 3.9% to 15.7% among traits and the genetic coefficient of variation (CVg) ranged from 2.2% to 10.0% (Table 2).The ratio Iv = CVg/CVe (Vencovsky, 1987;Vencovsky & Barriga, 1992) designated as index of variation (Valois, 1982) showed higher estimates (0.7 to 0.9) for PH, EH, TL, TW, TB, EL and RN, while smaller estimates (0.4 to 0.6) were found for EP, EW, GW, ED, KR and PR.Estimates of Iv for EW (0.7 and 0.5), PH (0.4 and 0.7) and EH (0.5 and 0.7) in two population (Miranda Filho & Gorgulho, 2003), were similar those showed in this study and also the estimates for a dwarf population in reduced spacing (Candido, 2005).However, estimates greater than one for this traits in dwarf population evaluated in density of 80,000 plants/ha were obtained for Garcia (2005).Higher Iv estimates indicate greater opportunities for selection, and Iv>1 would indicate a favorable condition for selection purposes (Vencovsky, 1987).
The estimates of the additive genetic variance for the tassel traits were 11.2 cm², 1.65 g² and 13.6 for TL, TW, and TB, respectively.The estimate for TB (13.6) was smaller than that reported (18.9) by Kassouf & Miranda Filho (1986) for ESALQ-PB1, but was greater than estimates for three different broad base populations reported by Geraldi et al. (1985).Additive variance estimates for TB varied from 4.0 to 53.1 in 40 estimates summarized by Araújo (1992), who found an estimate of 15.4 for ESALQ-PB1.For RN and KR the estimated additive variance were 0.018 and 10.9, respectively.
The estimates of the coefficients of heritability on an individual plant basis (Table 4) varied from 0.141 to 0.721 and were higher for PH, EH, and TB, intermediate for EP, TL, TW, EL, ED, and RN, and low for EW, GW, KR, and PR.The heritability estimates on progeny mean basis followed approximately the same trend and varied from 0.394 to 0.753, showing that some traits such GW and the yield components KR and PR can be more effectively selected on the basis of progeny means, while others exhibiting high individual heritability can be effectively selected on the basis of phenotypic expression.In general, the heritability estimates did not differ from estimates summarized by Geraldi et al. (1985), Miranda Filho (1985), Souza Jr. et al. (1985), Hallauer & Miranda Filho (1995)  ), coefficients of heritability (h 2 to individual and 2 m h to progeny mean basis) and expected gain from selection (G H to half-sib family selection and G M to mass selection) for the traits plant height (PH), ear height (EH), ear placement (EP), tassel length (TL), Prolificacy (PR), tassel weight (TW), tassel branch number (TB), ear weight (EW), grain weight (GW), ear length (EL), ear diameter (ED), kernel row number (RN) and kernels per row (KR), of the maize population ESALQ-PB1.
4, 5 -Estimates of 2 A σ multiplied by 10 4 and 10 5 , respectively.G H -10%: both sexes; G M -10% for both sexes for PH, EH, TB and PR and for female sex for other traits.greater were obtained by Candido (2005), Cárdenas (2005) and Garcia (2005).For GW, similar estimates were obtained in the BR 5011, CMS 35 and CPATC-3 brazilian populations (Carvalho et al., 2000(Carvalho et al., , 2002(Carvalho et al., and 2003)).All the estimates presented herein refer to only one environment (one location and one year) so that they may be overestimated by an amount proportional to the genotype by environment interaction when considering a sample of environments represented by several locations and/or years.
The expected gains from selection (Table 4) reflect, in most of the traits, their estimates of heritability; i.e., traits that exhibited higher heritability showed higher predicted gains for both mass selection and progeny selection.The greatest predicted gain from mass selection was for TB (27% per cycle) and from progeny selection was for TW (16.4% per cycle).The lowest expected gain was for ED from either mass selection (1.9%) or progeny selection (2.8%).When comparing mass selection with progeny selection on a per cycle basis, the mass selection was expected to be more effective for PH (11.3 vs. 5.8%), EH (14.1 vs. 7.8%) and TB (27.4 vs. 15.2%).PH, EH, and TB had the highest estimates for heritability coefficient and index of variation.The same comparison indicated progeny selection as more effective for TW (12.6 vs. 16.4%),EW (4.0 vs. 6.5%) and GW (4.2 vs. 7.0%).TW, GW, and EW had lower heritability estimates and are selected after flowering; i.e., for the female sex only.For the other traits, the expected gains showed smaller differences between selection schemes.
The expected gain for PH and EH in the base population ESALQ-PB1 indicated mass selection as more effective than selection among half-sib progenies.The same trends were not observed by other authors when considering selection within progenies in the recombination block (Ayala Osuna, 1977;Lordelo, 1982;Souza Jr., 1983;Soares Filho, 1987).For TB the efficiency of mass selection (27% per cycle) is comparable to that expected for the variety Centralmex (23.2% per cycle) and higher than that expected for Dent Composite (10.9%) and Flint Composite (13.4%) in which, however, a lower selection pressure (20%) was considered (Geraldi et al., 1985).Among the tassel traits, TB is indicated when selecting for tassel size due to its high heritability, and because it is easily evaluated in the field and highly correlated to TW (Geraldi et al., 1985).
The additive genetic correlation (r A ) and phenotypic correlation (r p ) on progeny mean basis (Table 5) were positive for most pairs of traits.PH and EH had r A between 0.48 and 0.60 with yield traits (EW and GW) and its components (EL and KR), which are in a higher level in relation to reports summarized by Hallauer & Miranda Filho (1995) as well as in relation to some Brazilian populations (Lima, 1977;Lima & Paterniani, 1977;Crisóstomo, 1978;Sampaio, 1986;Soares Filho, 1987).Positive correlations, however, do not preclude selection for pairs of traits in opposite direction to be effective.PH and EH were positively correlated to tassel traits with r A varying from 0.10 (EH.TL) to 0.44 (EH.TB).Also are reported positive genetic and phenotypic correlations between PH and EH with tassel traits (TB and TL) in the range of 0.135 to 0.438 (Farias Neto & Miranda Filho, 2001).For the tassel traits (Table 5) the highest correlation was for TB.TW (r A = 0.63).Correlation of TB and TW were positive and low (r A between 0.02 and 0.22) with the yield traits and higher (r A between 0.19 and 0.40) with EL.Both additive and phenotypic correlation between TW and TL was around 0.22, which are very close to the genetic correlation (0.232) reported by Farias Neto & Miranda Filho (2001).
The correlations indicate a low indirect effect on yield when selecting for tassel size.Although TB and TW are negatively correlated to PR (r A = -0.41 and r A = -0.48,respectively), PR showed no correlation (r A = 0.09) to the yield traits.High negative correlation between TB and PR in the maize population SUWAN and a high correlation (r A = 0.94) between GW and PR also were reported by Souza Jr. et al. (1985).Also estimated the expected correlated response (R s ) on yield as 11.9 and 16.6% when selecting (20% selection intensity) for decreasing TB and increasing PR, respectively.Our results indicated R s on yield of -2% (Table 6) when selecting (10% selection intensity) for decreasing TB, as a consequence of the low correlation between TB and GW in spite of the higher heritability for TB (Falconer & Mackay, 1996).Another important trait for the characterization of a better plant architecture in maize is EP, which was additively correlated with TB (0.27), TW (0.40), PH (0.20), EH (0.70), ED (-0.53),RN (-0.38), and PR (0.47).Although the correlation of EP with TB is not high (0.27), direct selection for decreasing TB will affect EP in the same direction; in the same way, selection for EH will affect both TB and EP.
The expected correlated responses in several traits through direct selection for GW, EH, and TB are shown in Table 6 for mass selection and progeny selection.If negative selection is practiced for TB or EH for better plant architecture, negative changes are expected in GW (2.0% and 4.7% respectively); however, the changes in GW can be minimized by using no truncated selection to avoid undesirable genotypes for yield.
A combined selection, including TB as selection criterion for three populations, based on the negative correlation between GW and TB to increase the gain in GW were suggested by Geraldi et al. (1985).The low correlation between GW and TB observed in ESALQ-PB1 makes the selection for TB to be feasible without great changes in the mean yield.In a progeny selection program, selection in ESALQ-PB1 should be performed among progeny means for yield and within progenies for EH and TB.In other populations and environments the expected correlated responses may be different as related by Garcia (2005) for dwarf maize in density of 80,000 plants/ha and reduced spacing.The expected correlated responses to family selection in GW was 8.8%, 15.1% and 9.5%, respectively for PH, EH and PR.With selection in EH, the expected correlated responses was 10.9%, 18.0% and 7.1%, for PH, GW and PR, respectively.The same tendency of larger correlated responses so that obtained in ESALQ-PB1, also was observed by Candido (2005).
The estimates of parameters related to genetic variability in the maize population ESALQ-PB1, shown in this study, refer to only one environment (one location and one year), so that a random model for environment cannot be considered and the genotype énvironment interactions are not taken into account.

Table 2 -
Mean, range (minimum and maximum), coefficients of variation (CV e : experimental; CV g : genetic) and index of variation (Iv) for 13 traits of the maize population ESALQ-PB1.Table 1 -Analysis of variance and covariance for individual experiments.MS and MP: mean squares and mean products; n, r, k: number of progenies, replications and plants per plot, respectively; k was not constant for all traits

April 2008 ing
estimates of variance components and other parameters (Table4).The estimates of the additive genetic variance for PH (309.3 cm²) and EH (175.0 cm²) did not differ greatly from the average (321.

Table 6 -
Expected correlated response on other traits after mass selection (M) and half-sib family selection (H) for grain yield (R G ), ear height (R E ), and tassel branch number (R T ) of the maize population ESALQ-PB1.