Reciprocal recurrent selection effects on the genetic structure of tropical maize populations assessed at microsatellite loci

A modified reciprocal recurrent selection (RRS) method, which employed one cycle of high-intensity selection, was applied to two tropical maize (Zea mays L.) populations, BR-105 and BR-106, originating the improved synthetics IG-3 and IG-4, respectively. In the present study the effects of this kind of selection on the genetic structure of these populations and their synthetics were investigated at 30 microsatellite (SSR) loci. A total of 125 alleles were revealed. A reduction in the number of alleles was observed after selection, as well as changes in allele frequencies. In nearly 13% (BR-105) and 7% (BR-106) of the loci evaluated, the changes in allele frequencies were not explained, exclusively due to the effects of genetic drift. The effective population sizes estimated for the synthetics using 30 SSR loci were similar to those theoretically expected after selection. The genetic differentiation (GST) between the synthetics increased to 77% compared with the original populations. The estimated RST values, a genetic differentiation measure proper for microsatellite data, were similar to those obtained for GST. Despite the high level of selection applied, the total gene diversity found in the synthetics allows them to be used in a new RRS cycle.


Introduction
Important increases in maize productivity have been obtained since the beginning of the last century because of the development of inbreeding and hybridization methods outlined by Shull (Crow, 1998).Currently, most maize breeding programs are based on hybrid production.The development of inbred lines and hybrids is very much related to the frequency of favorable alleles, which can be increased via recurrent selection (Hallauer and Miranda 1988).In this kind of selection, populations and inbred lines are developed to be crossed and to form superior hybrids.In the reciprocal recurrent selection method (RRS), genotypes from two populations are evaluated in reciprocal crosses, where each population is used as the other's tester.The improved populations are generated by intermating superior genotypes of each population that present the best combining abilities with the reciprocal population (Souza Jr., 1998).
Similarly to selection methods, RRS causes changes in the allele frequencies, levels and distribution of the ge-netic variability, and, consequently, in the genetic structure of the populations.The use of inadequate population sizes leads to the loss of genetic variability due to genetic drift effects.Such loss can limit long-term RRS programs (Guzman andLamkey, 1999, 2000).For this reason, highintensity selection has been avoided in conventional RRS.However, Rezende and Souza Jr. (2000) applied one cycle of high-intensity RRS in the tropical maize populations BR-105 and BR-106 and, despite the negative drift effects on the improvement of the populations per se, the interpopulation genetic variances were not significantly affected.
Molecular markers are promising for the investigation of all these changes.Labate et al. (1999) described significant changes in the allele frequencies at most maize loci after 12 cycles of RRS using the RFLP (restriction fragment length polymorphism) procedure, and genetic drift hypothesis was rejected by the Waples' neutrality test (Waples, 1989a).Koeyer et al. (2001) identified genomic regions containing favorable alleles using 97 RFLP loci to monitor genetic changes in a long-term recurrent selection program in oats.

Development of plant materials
Two maize populations, a Thai (BR-105) and a Brazilian composite (BR-106) and their synthetics were analysed.BR-105 and BR-106 were previously submitted to one cycle of high-intensity RRS (2.0 and 2.5%, respectively) using 400 S 3 lines.These lines were crossed with the opposite population and superior interpopulation half-sib (HS) progenies were identified.Eight S 3 lines (2.0% x 400) derived from BR-105 and ten S 3 lines (2.5% x 400) derived from BR-106, both related to the selected interpopulation HS progenies, were intercrossed in a diallel mating design within each population to develop IG-3 and IG-4 synthetics, respectively (Figure 1).The synthetics IG-3 and IG-4 resulted from intercrossing of S 3 lines i.e. they are the product of random mating of the alleles of eight lines derived from BR-105, and ten lines derived from BR-106, respectively.Considering that the effective population size (Ne) of each S 3 is approximately 0.57, the Ne of IG-3 and IG-4 are 4.56 and 5.71 with inbreeding coefficients of 10.9% and 8.75%, respectively (Rezende and Souza Jr., 2000).

DNA extraction and SSR assays
One hundred randomly chosen seeds were taken from each population and synthetics.Leaf tissues collected from 35 day-old plants were lyophilized, ground by a mechanical mill, and stored at -20 °C.Total genomic DNA was extracted from 300 mg of lyophilized tissues using a CTAB procedure (Hoisington et al., 1994).PCR reactions were performed in a 20 µL final volume containing 40 ng of template DNA, 0.2 µM of each forward and reverse primer, 100 µM of each dNTP, 2.0 mM MgCl 2 , 0.5 unit Taq DNA polymerase (Gibco-BRL), 10 mM Tris-HCl and 50 mM KCl.Reactions were run in a PTC-100 thermocycler (MJ Research) using the PCR cycling conditions described by Ogliari et al. (2000).Thirty primer pairs located at least in one maize chromosome were used to survey the genetic polymorphism.Amplification products were separated by electrophoresis on 3% agarose gels (50% agarose metaphor FMC-Bio products: 50% agarose Gibco-BRL) in TBE buffer (0,09 M Tris, 0,09 M boric acid, 2 mM EDTA).Gels were photographed under UV light after ethidium bromide staining.The sizes of the fragments were calculated by comparison with 50 and 100 bp ladders.

Allele frequencies
Individuals were genotyped in terms of their alleles and respective SSR loci, defined by a primer pair (forward/reverse).Allele frequencies were calculated using the BIOSYS-1 program (Swoffford and Selander, 1991).To test the hypothesis of identical distribution of the allele frequencies, an exact test for population differentiation was performed with the TFPGA program (Miller, 1997).The neutrality test (Waples, 1989a;Labate et al., 1999) was applied to each locus to verify whether the changes in allele frequencies after one cycle of RRS could be exclusively attributed to the genetic drift effects.

Diversity distribution
The distribution of gene diversity was conducted according to the model proposed by Nei (1973), in which the total genetic diversity mean (H T ) is partitioned in two components: the gene diversity mean within population (H S ), and between populations (D ST ).The proportion of total gene diversity (G ST ) between population, or genetic differentiation, was calculated as G ST = D ST /H T .To better understand the behavior of gene diversity after one cycle of selection, total gene diversity was performed separately for populations (before selection -C 0 ), and for synthetics (after selection -C 1 ), as well as for the combinations BR-105 vs. IG-3, and BR-106 vs. IG-4 using the FSTAT program (Goudet, 1995).Genetic differentiation was also estimated by using the R ST statistics, as the fraction of total variance in the allele size (in base pairs) that occurs between populations, using the RSTCALC package (Goodman, 1997).
Effective population size IG-3 and IG-4 effective population sizes (Ne) were estimated according to the Waples method (1989b), based on Plan II, where the individuals are taken before the reproduction event, and not replaced.Considering N 0 and N t as the respective sampling sizes at the two sampling events, t the time between the two sampling events, and $ Fc the weighted standardized variance in allele frequencies, Ne is given by The standardized variance in allele frequencies for each locus ( $ Fc) was calculated using the expression proposed by Nei and Tajima (1981): where p u and p' u are the frequencies of the u allele at the two sampling events and k is the number of alleles at a locus.

For multiple loci,
$ Fc is given by the weighted means of single locus values ( $ Fc) by the number of alleles at each locus.The 95% confidence intervals were calculated using the formula: based on the number of k independent alleles ( ( )

Allele frequencies
The 30 loci revealed a total of 125 alleles, 111 occurring in BR-105 and 116 in BR-106 (Table 1).Most of the alleles that were in low frequency in the original populations were lost after one cycle of RRS.Allele reductions were observed in IG-3 (23%) and IG-4 (17%).An increase in the number of alleles belonging to the extreme classes of frequencies was detected in both populations after selection (Figure 2).This is a feature of a dispersive process in which the allele frequencies tend towards the limits of zero (lost) or 1 (fixation).The differentiation tests for the allele fre-Genetic structure of maize populations 357  C 1 (G ST = 19.5%)revealed an increase of 77.3%.Consequently, the synthetics became more divergent.Such allele losses contributed for this differentiation.
The genetic differentiation (G ST ) for the combinations BR-105 vs. IG-3, and BR-106 vs. IG-4 was greater for the first group (12.4% vs. 6.8%) in which the selection intensity was higher (2.0% vs. 2.5%).These G ST values were statistically significant.The R ST values were slightly superior to the G ST values.As pointed out by Gaiotto et al. (2001) the R ST statistic can be used as evidence of genotyp-Genetic structure of maize populations 361  ing accuracy due to the fact that its estimation is based on the magnitude of the variances.

Effective population size
The values for the effective population sizes estimated for the synthetics IG-3 (3.87) and IG-4 (6.62) were similar to those theoretically expected, i.e. for the recombination of 8 and ten S 3 lines, respectively (Table 5).In practical terms, it shows that the samples (on average 93 individuals) represent approximately 3.87 and 6.62 plants of an ideal panmitic population and correspond to 4.16% and 7.11% of the total sampled individuals from IG-3 and IG-4, respectively.The estimated effective population sizes provided inbreeding coefficients [(F = 1/2Ne).100] of 12.91% for IG-3 and 7.55% for IG-4, which are similar to the expected values of 10.94% and 8.75%, respectively.

Discussion
The differences in the allele frequency distributions between the original populations supported the existence of genetic divergence reported by Naspolini Filho et al. (1981) because of the magnitude of the heterosis manifested in the interpopulation cross.
Changes in allele frequencies between populations and respective synthetics were observed by the lack of overlapping of the confidence intervals in nearly 50% of the alleles.At most of the loci, the nature of these changes was due to stochastic processes i.e., genetic drift and sampling errors.Nevertheless, changes at some loci were highly significant, and therefore were not due to effects of drift alone.
The apparent lack of neutrality in RRS programs was verified by Labate et al. (1999) at 17% of the RFLP loci dis-persed in the maize genome.This result was interpreted as a selection by genetic hitchhiking.As microsatellite loci represent repetitive and non-coding DNA regions, they apparently are not subject to strong selection pressures (Heath et al., 1993).However, they can be linked to selected loci, and therefore subjected to selection by genetic hitchhiking.
The loss of total gene diversity detected at microsatellite loci (9.4%) was similar to the decrease of genetic variance (8.8%) obtained for yield (Table 6).A similar result was observed in the BSSS and BSCB1 maize populations in which the loss of genetic diversity assessed at RFLP markers was consistent with the decrease in additive and dominant genetic variance in the BSSS population after 12 RRS cycles (Holthaus and Lamkey, 1995;Labate et al., 1999).Both studies confirmed that diversity, or expected heterozygosity, is proportional to genetic variance (Lacy, 1987).
Despite the intensity of the applied selection, our results showed that the total gene diversity loss was not so large after one cycle of RRS.According to Rezende and Souza Jr. (2000), high-intensity RRS selection in BR-105 and BR-106 caused a significant increase in heterosis 362 Pinto et al.   (25.7% for grain yield) and improved the cross performance between the synthetics IG-3 and IG-4, but the interpopulation genetic variances did not present significant changes (Table 6).The maintenance of genetic variability has been reported in several long and short-term recurrent selection programs (Bernardo, 1996;Guzman and Lamkey, 1999), and by Labate et al. (1999), using RFLP markers.In addition, the use of large effective population sizes has shown no advantage for maintaining genetic variability for short-term recurrent selection (Guzman and Lamkey, 2000).The increase in the genetic differentiation (G ST ) after selection (77.3%) i.e., in the genetic divergence between the synthetics is predictable in a RRS program.The maintenance of two separate genetic pools allows different alleles to be fixed in each population and guarantees the heterozygous condition for these loci in the interpopulation hybrids.This situation can maximize heterosis whose expression considers the genetic divergence between populations and the dominance effects attributed to the heterozygous loci (Lamkey and Edwards, 1999).
In our results, microsatellites proved to be very promising for monitoring genetic variability and gave support for the use of the improved synthetics in a next cycle of selection.In agreement with previous studies, short-term recurrent selection does not require large effective population sizes, once the genetic variability is maintained at adequate levels after a high-intensity RRS procedure.We can conclude that the modified RRS process, here investigated, would successfully replace the traditional RRS procedure in maize.

Figure 1 -
Figure 1 -Description of the modified RRS in which a high-intensity of selection was applied to BR-105 and BR-106 giving rise to the synthetics IG3 and IG4.

Table 2 -
Differences in the allele frequency distribution in the four combinations between maize materials.

Table 3 -
Neutrality test for changes in allele frequencies after one cycle of RRS in the BR-105 and BR-106 maize populations.

Table 4 -
Mean values of gene diversity between the original populations BR-105 and BR-106 and their synthetics IG-3 and IG-4 after one cycle of RRS.