Published in Crop Sci. 44:401-404 (2004).
© 2004 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
CROP BREEDING, GENETICS & CYTOLOGY
Random Mating before Selfing in Maize BC1 Populations
Martin Arbelbide and
Rex Bernardo*
Dep. of Agronomy and Plant Genetics, Univ. of Minnesota, 411 Borlaug Hall, 1991 Upper Buford Circle, St. Paul, MN 55108-6026
* Corresponding author (berna022{at}umn.edu).
 |
ABSTRACT
|
|---|
In maize (Zea mays L.) inbred development programs, F2 and BC1 populations between two inbreds are usually not random-mated before selfing. Previous results suggest random mating is not useful in F2 populations. A one-locus model, however, suggests random mating increases the genetic variance by 50% in a BC1 population without affecting the population means. The objective of this study was to determine the effect of random mating on the testcross means and variances of BC1 populations. Random-mated and nonrandom-mated backcross populations were developed for two genetic backgrounds, Iowa Stiff Stalk Synthetic (BSSS) and non-BSSS. Testcrosses of these four populations were evaluated at five locations in Minnesota and Wisconsin in 2002. For grain yield, grain moisture, plant height, ear height, root lodging, and stalk lodging, the differences between testcross means of random-mated and nonrandom-mated populations were not significant (P = 0.05), as expected. Testcross variances tended to increase with random mating for most traits in the two genetic backgrounds, but none of these differences was significant. Testcross means of the best 10% families did not differ significantly between random-mated and nonrandom-mated populations. Overall, the results indicated that random mating before selfing in BC1 populations is not useful in applied breeding programs.
Abbreviations: BSSS, Iowa Stiff Stalk Synthetic • REML, restricted maximum likelihood • VTC, testcross variance
|
INTRODUCTION
|
|---|
THE TWO MOST common breeding populations used in inbred development are F2 and BC1 populations from a cross between two inbreds. The choice between an F2 and BC1 population depends on breeding objectives and the relative performance of the parental inbreds (Dudley, 1982; Bernardo, 2002, p. 77). A BC1 population is more desirable when one parental inbred is superior to the other, when short-term response is desired, or when selection intensity is low (Dudley, 1982; Bridges and Gardner, 1987; Lamkey et al., 1995).
An important question in inbred development is whether recombination by random mating before inbreeding and selection is advantageous (Hallauer, 1990). Maize F2 and backcross populations are generally not random-mated in breeding programs. Random mating an F2 population has been shown to be of little advantage (Altman and Busch, 1984; Covarrubias-Prieto et al., 1989; Han and Hallauer, 1989; Lamkey et al., 1995), but a one-locus model suggests that random mating might be useful in a BC1 population.
Consider two inbred parents, Parent 1 with genotype A1A1, Parent 2 with genotype A2A2, and an inbred tester with genotype ATAT, for a given trait (Table 1). The BC1 is created by crossing Parent 1 with Parent 2, and backcrossing the F1 (A1A2) to Parent 1. The resulting BC1 population is composed of 1/2 A1A1 and 1/2 A1A2 individuals. If the BC1 population is random-mated, the genotypic frequencies become 9/16 A1A1, 6/16 A1A2, and 1/16 A2A2. Random mating, therefore, changes the frequencies of the A1A1 and A1A2 genotypes and permits the recovery of the donor parent genotype, A2A2, at low frequency. These changes in genotypic frequencies do not affect testcross means, but they increase VTC by 50% (Table 1). Testcross genotypic values are additive regardless of the level of dominance (Hallauer and Miranda, 1988, p. 28; Bernardo, 2002, p. 79), so these results hold for any level of dominance. The increase in VTC could then lead to a larger response to selection.
View this table:
[in this window]
[in a new window]
|
Table 1. Genotypes, genotypic frequencies, and testcross genotypic values in a BC1 population for a one-locus model.
|
|
No research has been conducted to address the usefulness of random mating before selfing in backcross populations between elite inbreds. The objective of this study was to determine the effect of random mating on the testcross means and variances of BC1 populations and to assess whether results fit the expectations for a one-locus model.
|
MATERIALS AND METHODS
|
|---|
We used four populations developed by Syngenta Seeds, Inc. (Stanton, MN): (i) a BSSS BC1 population, (ii) a random-mated BSSS BC1 population (BC1–Syn1), (iii) a non-BSSS BC1 population, and (iv) a random-mated non-BSSS (BC1–Syn1) population. The BSSS BC1 and non-BSSS BC1 were derived from crosses between two inbreds. The BSSS BC1–Syn1 was created by random mating the BSSS BC1 population, and the non-BSSS BC1–Syn1 population was created by random mating the non-BSSS BC1 population. Random mating was done by crossing pairs of plants in each BC1 in St. Paul, MN, in the summer of 2001. About 320 plants were random-mated in each BC1 with the restriction that a plant could be used only once, as either the female or male parent. The BSSS BC1 and BSSS BC1–Syn1 populations were testcrossed to a non-BSSS inbred, and the non-BSSS BC1 and non-BSSS BC1–Syn1 populations were testcrossed to a BSSS inbred. Testcrosses were made in Hawaii in the 2001-2002 winter season. BC1 and BC1–Syn1 plants were used as males and tester plants used as females. Each male plant was testcrossed to four female plants.
Yield trials were conducted in the summer 2002 season at five locations: Rosemount, Waseca, and Lamberton, MN, and Janesville and Madison, WI. The soil types at these locations were Waukegan silt loam (Rosemount; fine-silty over sandy or sandy-skeletal, mixed, superactive, mesic Typic Hapludolls), Webster clay loam (Waseca; fine-loamy, mixed, superactive, mesic Typic Endoaquolls), Normania loam (Lamberton; fine-loamy, mixed, superactive, mesic Aquic Hapludolls), and Plano silt loam (Janesville and Madison; fine-silty, mixed, superactive, mesic Typic Argiudolls). A sets-in-blocks design (Comstock and Robinson, 1952) with a single replication was used at each location. For the BSSS populations, 150 BC1 and 150 BC1–Syn1 families were randomly divided into 10 sets. For the non-BSSS populations, 210 BC1 and 210 BC1–Syn1 families were randomly divided into 14 sets. Two-row plots, 6.7 m (MN) or 6.1 m (WI) long and spaced 0.76 cm apart, were used in this study. Plots were sown for a mean plant population density of 82000 plants ha–1. Grain yield (Mg ha–1) was measured as the harvested grain weight adjusted to 155 g H2O kg–1. Grain yield and moisture (g H2O kg–1) were recorded with a plot combine. Plant height in centimeters was measured on a plot basis (i.e., visually determined mean for the plot) as the distance from the soil surface to the tip of the tassel. Ear height in centimeters was measured on a plot basis as the distance from the soil surface to the node of insertion of the top ear. Percentage root lodging was scored as the number of plants leaning at an angle lower than 45 degrees with respect to the soil surface divided by the total number of plants in the plot. Grain yield and moisture were recorded at all five locations. Plant and ear height were recorded at three locations (Lamberton, Rosemount, and Waseca, MN). Root lodging was recorded at two locations (Janesville and Madison, WI).
The four populations were analyzed separately. Testcross genetic variance components were estimated by ANOVA and restricted maximum likelihood (REML), pooled across sets, and combined across locations. Genotypes, locations, and sets were considered random effects. Mean squares and variance components were estimated with the VARCOMP procedure of SAS (SAS Institute, 1990). The ANOVA in VARCOMP allowed the estimation of mean squares for the genotypic and error effects. The REML analysis in VARCOMP provided estimates of the asymptotic variance of the VTC estimates. Confidence intervals (P = 0.95) were calculated for the means. Ratios between SEs and means were calculated as an indication of the precision in the estimation of means (Knapp et al., 1987). Confidence intervals (P = 0.90) for VTC were calculated as described by Knapp et al. (1987). The ratio of VTC in the random-mated and the nonrandom-mated populations (VBC1–Syn1/VBC1) was estimated. The variance of the ratio VBC1–Syn1/VBC1 was estimated as described by Lynch and Walsh (1998)(p. 818). Testcross heritability (h2) was estimated on a testcross-mean basis as the ratio between VTC and the phenotypic variance of a testcross mean. Means of the best 10% families for each trait were calculated for the four populations. Differences between the means of the best 10% families in the BC1–Syn1 and BC1 populations were tested with a two-sample t test (P = 0.05).
|
RESULTS AND DISCUSSION
|
|---|
Testcross Means
In both the BSSS and non-BSSS populations, the testcross means for all traits studied were not significantly different before and after random mating (Table 2). Grain yield differences (0.14 Mg ha–1) attributable to random mating the non-BSSS population were statistically significant (P < 0.05) but not agronomically significant. The ratios between SEs and means were low for grain yield (0.77%), grain moisture (0.36%), plant height (0.31%), and ear height (0.61%), but not for lodging (14%), where data from only two locations were available. These results indicated that the estimates of overall testcross means were precise.
As previously mentioned, the one-locus model indicates that testcross means are not expected to change with random mating in the absence of selection (Table 1). When a two-locus model with arbitrary linkage is considered, testcross means are also not expected to change after random mating unless selection or epistasis are present. Hoffbeck et al. (1995) found no changes in population means after exotic x adapted maize backcross populations were intermated for five generations. Results indicated that population means after random mating were not affected by unintentional selection and epistatic effects. Most studies on the effects of random mating were conducted on F2 populations and involved more than one generation of random mating. Covarrubias-Prieto et al. (1989) reported that means of F2 populations are also not expected to change after random mating if selection and epistatic effects are absent. In contrast, they found small changes in the means for grain yield and grain moisture before and after six generations of random mating maize F2 populations. Differences were attributed to epistatic effects or natural selection during the six generations of random mating. Similarly, Lamkey et al. (1995) found a significant reduction in grain yield after five generations of random mating F2 populations. Differences were attributed to recombination and break-up of favorable allele combinations (Melchinger et al., 1988). In other studies (Miller and Rawlings, 1967; Humphrey et al., 1969; Meredith and Bridge, 1971; Altman and Busch, 1984), changes in population means were attributed to epistatic effects or natural selection. Unintentional selection is likely to become important after many generations of random mating. In our study, testcross means were unaffected by random mating, suggesting that epistatic effects were negligible, and with only one generation of random mating, natural selection was absent. Our results for testcross means provide evidence in support of a one-locus model.
Testcross Variances
Most estimates of VTC were different from zero (Table 3). For grain yield, however, VTC was different from zero only in the BSSS BC1–Syn1 population. Estimates of VTC for root lodging were not significantly different from zero in all populations. Testcross heritability was highest for grain moisture and ear height and lowest for root lodging.
As previously mentioned, a one-locus model shows that random mating a BC1 population is expected to increase VTC by 50%; that is, VBC1–Syn1/VBC1 = 1.5. Most traits (five of eight) showed a VBC1–Syn1/VBC1 ratio larger than 1.0 (Table 4). The VBC1–Syn1/VBC1 ratio for grain yield in the BSSS background was not calculated because of a negative VTC estimate. Grain yield had a ratio lower than expected (0.81) in the non-BSSS background. The largest VBC1–Syn1/VBC1 ratios were for grain moisture (1.91) and ear height (1.88) in the BSSS background. The VBC1–Syn1/VBC1 ratio for root lodging was lower than expected in the BSSS population. Plant height had a lower ratio in the BSSS than in the non-BSSS backgrounds, and in both cases, ratios were lower than expected. The mean ratio for all traits was 1.29 for the BSSS and 1.05 for the non-BSSS populations. In other words, the mean increase in genetic variance was lower than expected in both the BSSS and the non-BSSS populations. Ratios higher or lower than 1.5 cannot be explained by a one-locus model. Considering the standard deviations of the VBC1–Syn1/VBC1 ratios, however, the estimates of the VBC1–Syn1/VBC1 ratio were not statistically different from either 1.0 or 1.5. Overall, random mating seemed to increase genetic variances, though the estimates were not precise enough to achieve statistical significance. Evidence for or against a one-locus model was inconclusive. Our preliminary investigations of a two-locus model indicated that if epistasis and linkage are present, random mating would lead to a VBC1–Syn1/VBC1 ratio of about 1.6 instead of 1.5. Therefore, the difference between a one-locus and two-locus model seemed minor, and our results were not precise enough to distinguish one model from the other.
Hoffbeck et al. (1995) intermated adapted x exotic maize BC1 populations for five generations and found no effects of random mating on genetic variances. The lack of change in genetic variances after random mating was attributed to insufficient genetic recombination. Most studies on the effects of random mating with inbred parents have been conducted with F2 populations. According to a one-locus model, random mating is not expected to affect genetic variance in an F2 population in the absence of selection. In the study by Covarrubias-Prieto et al. (1989), some populations had significant changes in genetic variances while others were not affected. Results also varied among traits. Changes were attributable to recombination, linkage disequilibrium, or genetic drift during random mating. Similarly, Han and Hallauer (1989) found no changes in additive genetic variance before and after five generations of random mating two F2 populations. Estimates of dominance variance decreased with random mating, implying the presence of linkage effects. Lamkey et al. (1995) found no increases in genetic variance after random mating F2 populations for five generations. Results varied among traits, but grain yield was the most affected trait. Covarrubias-Prieto et al. (1989), Han and Hallauer (1989), Hoffbeck et al. (1995), and Lamkey et al. (1995) all concluded that random mating had little effect on genetic variances, and recombination before selfing and selection is not recommended in F2 populations.
Implications for Applied Breeding Programs
Applied breeders are only indirectly interested in genetic variance; they are directly interested in the mean of the best testcrosses. The mean of the best 10% of testcrosses combines information on the mean and variance of the population. If VTC in the BC1 population is 1.0 and testcross performance is expressed as a deviation from the population mean, then the mean of the best 10% testcross families would be µ10% = k10%
= 1.76 (
) = 1.76, where k10% is the standardized selection differential for a selection intensity of 10%. Given that random mating is expected to increase VTC by 50%, the mean of the best 10% of testcross families in the BC1–Syn1 population would be µ10% = 1.76(
) = 2.15. The mean of the best 10% testcross families would, therefore, increase by 22% (from 1.76 to 2.15) after random mating. In our study, however, there were no significant differences in the means of the best 10% of testcrosses before and after random mating (Table 2). These results applied to traits with low h2 (grain yield and root lodging) as well as to traits with moderate h2 (grain moisture, ear height, and plant height). Random mating seemed to increase genetic variance, but the increase was not sufficient to have an effect on the means of the best testcrosses. Random mating tended to skew trait distributions toward the lower tails, but differences in skewness or kurtosis of the BC1 and BC1–Syn1 populations were not significant (results not shown).
Random mating is seldom used by plant breeders because the breeding process is lengthened by one generation, which breeders would rather use for additional selection (Bernardo, 2002, p. 24). Random mating must prove advantageous by increasing the genetic variance in the breeding population, and it must also outweigh the time and resources it requires. In our study, we conducted a large experiment (150 to 210 testcrosses per population) to estimate changes in VTC by random mating in two BC1 populations. In contrast, a commercial maize breeder typically develops 50 to 100 families in each breeding population (Bernardo, 2002, p. 13). The use of small breeding populations may prevent the breeder from taking advantage of a 50% increase in VTC upon random mating. In other words, if the effects of random mating were not evident in our study (which utilized large numbers of testcrosses), then the effects of random mating are unlikely to be evident in a breeding program with far fewer testcrosses per population. The results of our study indicated that the increase in VTC may not be sufficient to justify the time and resources that random mating requires. We conclude that random mating before selfing in BC1 populations is not useful in applied breeding programs.
|
ACKNOWLEDGMENTS
|
|---|
We thank Brent Delzer and Steve Damon from Syngenta Seeds for providing the germplasm as well as nursery and field support for this study.
Received for publication May 10, 2003.
|
REFERENCES
|
|---|
- Altman, D.W., and R.H. Busch. 1984. Random intermating before selection in spring wheat. Crop Sci. 24:1085–1089.[Abstract/Free Full Text]
- Bernardo, R. 2002. Breeding for quantitative traits in plants. Stemma Press, Woodbury, MN.
- Bridges, W.C., and C.O. Gardner. 1987. Foundation populations for adapted by exotic crosses. Crop Sci. 27:501–506.[Abstract/Free Full Text]
- Comstock, R.E., and H.F. Robinson. 1952. Estimation of the average dominance of genes. p. 494–516. In J.W. Gowen (ed.) Heterosis. Iowa State Univ., Ames.
- Covarrubias-Prieto, J., A.R. Hallauer, and K.R. Lamkey. 1989. Intermating F2 populations of maize. Genetika 21:111–125.
- Dudley, J.W. 1982. Theory for transfer of alleles. Crop Sci. 22:631–636.[Abstract/Free Full Text]
- Hallauer, A.R. 1990. Methods used in developing maize inbreds. Maydica 35:1–16.
- Hallauer, A.R., and J.B. Miranda Fo. 1988. Quantitative genetics in maize breeding. 2nd ed. Iowa State Univ. Press, Ames.
- Han, G.-C., and A.R. Hallauer. 1989. Estimates of genetic variability in F2 maize populations. J. Iowa Acad. Sci. 96:14–19.
- Hoffbeck, M.D., S.J. Openshaw, J.L. Geadelmann, R.H. Peterson, and D.D. Stuthman. 1995. Backcrossing and intermating in an exotic x adapted cross of maize. Crop Sci. 35:1359–1364.[Abstract/Free Full Text]
- Humphrey, A.B., D.F. Matzinger, and C.C. Cockerham. 1969. Effects of random intercrossing in a naturally self-fertilizing species, Nicotiana tabacum L. Crop Sci. 9:495–497.[Abstract/Free Full Text]
- Knapp, S.J., W.M. Ross, and W.W. Stroup. 1987. Precision of genetic variance and heritability estimates from sorghum populations. Crop Sci. 27:265–268.[Abstract/Free Full Text]
- Lamkey, K.R., B.J. Schnicker, and A.E. Melchinger. 1995. Epistasis in an elite maize hybrid and choice of generation for inbred development. Crop Sci. 35:1272–1281.[Abstract/Free Full Text]
- Lynch, M., and B. Walsh. 1998. Genetics and analysis of quantitative traits. Sinauer Assoc., Sunderland, MA.
- Melchinger, A.E., W. Schmidt, and H.H. Geiger. 1988. Comparison of testcrosses produced from F2 and first backcross populations in maize. Crop Sci. 28:743–749.[Abstract/Free Full Text]
- Meredith, W.R., and R.R. Bridge. 1971. Breakup of linkage blocks in cotton, Gossypium hirsutum L. Crop Sci. 11:695–698.[Abstract/Free Full Text]
- Miller, P.A., and J.O. Rawlings. 1967. Breakup of initial linkage blocks through intermating in a cotton breeding population. Crop Sci. 7:199–204.
- SAS Institute. 1990. SAS/STAT user's guide. v. 6. 4th ed. Vol. 2. SAS Inst., Cary, NC.
TOP
ABSTRACT
INTRODUCTION
