Crop Science Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 23 September 2005
Published in Crop Sci 45:2301-2306 (2005)
© 2005 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Tabanao, D. A.
Right arrow Articles by Bernardo, R.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Tabanao, D. A.
Right arrow Articles by Bernardo, R.
Agricola
Right arrow Articles by Tabanao, D. A.
Right arrow Articles by Bernardo, R.
Related Collections
Right arrow Crop Genetics
Right arrow Crop Models

CROP BREEDING, GENETICS & CYTOLOGY

Genetic Variation in Maize Breeding Populations with Different Numbers of Parents

Dindo A. Tabanao 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 (bernardo{at}umn.edu)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
The development of breeding populations from two related inbreds restricts the amount of variability expected in the cross. Our objective was to determine if, as expected from an additive genetic model, genetic variance in maize (Zea mays L.) decreases as the number of parents of the population decreases. Eight random F2 plants derived from the same single cross were selected as parents and were selfed to form F3 families. These F3 families were intermated in a hierarchical manner to form populations with N = 1, 2, 4, or 8 parents. Testcross families were evaluated in four Minnesota environments in 2003. Marker heterozygosity at 62 simple sequence repeat loci decreased with N and had a near-linear association with expected heterozygosity. Testcross genetic variance (VTC) generally decreased with N, although differences observed at N = 2, 4, and 8 were mostly insignificant. For grain yield, VTC was significantly higher at N = 4 than at N = 2. The VTC for grain yield at N = 4 was slightly higher than the estimate of VTC in the base population, strongly indicating a genetic variance in excess of what is predicted by an additive genetic model for a small population. Differences in testcross means were likewise inconsistent with an additive genetic model. The results imply that nonadditive gene effects in elite maize inbreds help maintain genetic variance in small populations and that using multiple parents may help sustain genetic variability in advanced cycle breeding.


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
BREEDING POPULATIONS in different crops are most commonly created from crossing two inbred parents. Crossing few parents limits the genetic variability expected in the population. If the additive genetic variance in a large population is VA, the expected genetic variance of a purely additive trait in a small population is (1 – 1/2N)VA, where N is the number of noninbred parents (Lande, 1980). If the parents are inbred or related to each other, the expected variability in the cross becomes even less.

The narrowing genetic base is a main cause of potential yield plateaus in major crops, such as cotton (Gossypium hirsutum L., Meredith, 2000), rice (Oryza sativa L., Peng et al., 1999), and wheat (Triticum aestivum L., Araus et al., 2002). Reports involving bottlenecks in other organisms have indicated, however, that genetic variation does not always decrease with small population size (Bryant et al., 1986; Cheverud and Routman, 1996). This phenomenon has been demonstrated in fitness-related traits in many animal species, including Musca domestica (Bryant et al., 1986), Drosophila melanogaster (López-Fanjul and Villaverde, 1989), Bicyclus anynana (Saccheri et al., 1996), Tribolium castaneum (Wade et al., 1996), and Mus musculus (Cheverud et al., 1999). Epistasis has been shown to theoretically explain why bottlenecked populations exhibit more variation than expected under an additive genetic model (Goodnight, 1987, 1988, 2000). The VA, dominance variance (VD), or both VA and VD may increase under inbreeding if certain types of epistasis are present (Cheverud and Routman, 1996; López-Fanjul et al., 1999, 2002; Naciri-Graven and Goudet, 2003).

Population geneticists study bottlenecks in the laboratory by extracting a few pairs of individuals from a base population and letting them intermate to create a new population. Plant breeders enforce bottlenecks when they select only a fraction of the extensive array of germplasm available—both within and across breeding populations and programs—to constitute parents for the next generation. The number of parents selected in plant breeding therefore reflects effective population size in natural populations. If epistasis has been implicated in the excess of genetic variation in inbreeding natural populations, then it might also be causing some of the genetic variation in closely related elite x elite crosses in crops. By varying the number of parents used to form a breeding population to enforce four different levels of inbreeding, we aimed to determine if, as expected from an additive genetic model, genetic variance in a maize breeding population decreases as the number of parents decreases.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 REFERENCES
 
Development of Test Populations
We developed an F2 population from the cross between two inbreds that had good combining ability when crossed to Iowa Stiff Stalk Synthetic (BSSS) inbreds. The two parental inbreds were commercial inbreds licensed from a private maize breeding program. Eight F2 plants (designated A through H) were randomly selected as parents and were selfed to obtain F3 families. These F3 families were crossed in a hierarchical mating scheme to produce populations with different N in summer 2001 in Saint Paul, MN. The following intercrosses were performed: AxA, BxB, ..., HxH (to form eight N = 1 populations); AxB, CxD, ExF, GxH (to form four N = 2 populations); AxBxCxD, ExFxGxH (to form two N = 4 populations); and AxBx...xH (to form one N = 8 population). To develop each population, seeds from the appropriate F3 families were mixed in equal amounts, and crosses between random pairs of plants were made. This intermating served as the first generation of random mating. A second generation of random mating was performed in winter 2001-2002 in Puerto Rico through paired-row crosses within populations. In summer 2002 at Saint Paul, S0 plants from each population were testcrossed (as female parents) to BSSS inbred LH295 (male parent) in an isolated field. Ears were harvested separately for each S0 plant to form testcross families. The number of families for each population ranged from 30 to 200, for a total of 840 testcross families (Table 1).


View this table:
[in this window]
[in a new window]
  		Table 1. Number of families in maize testcross populations with different numbers of parents.

 
Field Evaluation
In summer 2003, the 840 testcross families were evaluated in a replications-in-sets incomplete block design at four Minnesota locations: Lamberton, Madison, Rosemount, and Waseca. The families from each population were equally divided among 10 sets (i.e., 3 from each N = 1 population, 5 from each N = 2 population, 10 from each N = 4 population, and 20 from the N = 8 population). Each set was replicated twice. Each plot consisted of two rows 5.3 to 6.7 m long at different locations and spaced 0.76 m apart. Plant population densities (plants ha–1) were 68700 at Lamberton and Waseca, 75600 at Rosemount, and 79000 at Madison. Ear height (distance in cm from soil surface to node of top ear), plant height (distance in cm from soil surface to tip of tassel), grain moisture (g H2O kg–1), and grain yield (Mg ha–1, adjusted to 155 g H2O kg–1) were measured on each plot. Ear height and plant height were recorded at Madison, Rosemount, and Waseca. Grain moisture and grain yield were recorded at all four locations.

Statistical Analysis
Phenotypic data were analyzed by SAS/STAT (SAS Institute Inc. 1999). Variance components were estimated from mean squares calculated through general linear models in PROC GLM. Each trait was modeled as a function of the following effects: locations, sets nested in locations, location-set interaction, replications nested in location-set interaction, populations, location-population interaction, location-set-population interaction, testcrosses nested in set-population interaction, and location-testcross interaction nested in set-population interaction. Locations and testcrosses nested in set-population interaction were considered to be random effects.

Estimates of testcross means and testcross genetic variances (VTC) were obtained within each level of N. Mean squares (MS) were obtained from sums of squares pooled from populations with the same N. The VTC was estimated as VTC = (MSTestcross – MSLocationxTestcross)/lr, where l was the number of locations and r was the number of replications. Based on the additive testcross model, VTC = (1 – 1/2N)VTC(N = {infty}), VTC at different N was fitted without intercept in a least-squares linear regression on 1/2N to estimate VTC(N = {infty}), which was the testcross genetic variance in the base population. The VTC(N = {infty}) was then used to predict the expected VTC for comparison with the estimates resulting from the field experiments. Approximate 90% confidence intervals for the estimates of VTC were constructed following the method described by Knapp et al. (1987). The VTC was F-distributed with degrees of freedom for testcross and location x testcross sources of variation. Significant differences between VTC estimates from different N were determined by Satterthwaite approximate F tests (Oehlert, 2000, p. 262). The usefulness criterion (Up) was calculated to predict the mean yield after selecting the upper proportion p, considering both mean and variance, of a population. The Up = µ + kpVTCV–1/2P, where µ was the mean of the testcross population, kp was the standardized selection differential, and VP was the phenotypic variance (Melchinger et al., 1988). Three selection intensities were used: 1, 10, and 20%.

SSR DNA Assay
Leaf tissue was collected and bulked from each of the eight N = 1 populations in summer 2002. Genomic DNA was extracted by the CTAB method (Kidwell and Osborn, 1992). Polymerase chain reaction (PCR) was performed in a GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA) thermocycler. The following temperature profile was used: 9-min initial denaturation at 95°C, followed by 35 cycles of 1 min denaturation at 95°C, 2 min annealing at 52, 55, or 58°C, and 2 min extension at 72°C; then a 7-min final extension at 72°C. Each PCR mix (15 µL) consisted of the following: 1.0 µL 10x PCR buffer, 0.8 µL 25 mM MgCl2, and 0.05 µL 5 U/µl DNA polymerase from the AmpliTaq Gold kit (Applied Biosystems, Foster City, CA); 1.5 µL 1.25 mM GeneAmp dNTPs (Applied Biosystems, Foster City, CA), 0.1 µL 10 µM forward primer, 0.1 µL 10 µM reverse primer, and 4.45 µL HPLC water. PCR products were run in 4% (w/v) denaturing polyacrylamide gel and visualized by silver stain. Bands were scored as codominant SSR alleles.

Because each DNA sample was a bulk from many plants in a population, the observed allele type(s) in each SSR locus automatically became the allele type(s) of the population at that locus. Average marker heterozygosity (HM) for each population was therefore calculated as the number of loci at which the population was heterozygous, divided by the total number of polymorphic loci. This statistic was based on Weir's (1996)(p. 141) measure of heterozygosity, except that in this study, a population was represented by a bulked DNA sample from many plants.

A total of 154 maize SSR markers (http://www.maizegdb.org/ssr.php, accessed 10 June 2005) were used in PCR runs, but only 62 markers that were polymorphic and had good amplification signals were used in the analysis. A list of all the primers assayed is available from the authors. The allelic arrays of the N = 2, 4, and 8 populations were based on the allelic bands of their corresponding parents, which were the N = 1 populations.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
REFERENCES
 
Genetic Variance
The testcross genetic variance in the base population [VTC(N = {infty}), shown as dotted line in Fig. 1] was 0.14 Mg2 ha–2 for grain yield, 78.46 g2 kg–2 for grain moisture, 37.08 cm2 for plant height, and 18.70 cm2 for ear height. Grain yield VTC at N = 4 (0.15 Mg2 ha–2) slightly exceeded VTC(N = {infty}) (0.14 Mg2 ha–2), and ear height VTC at N = 2 (18.03 cm2) was almost equal to VTC(N = {infty}) (18.70 cm2). The unexpected amount of VTC in these two instances suggests a release of VA because of fixation of epistatic loci (Goodnight, 1988). The increase in VTC for grain yield at N = 4 also suggests that a population may not experience any loss at all in genetic variance when it undergoes a bottleneck. The occurrence of the highest VTC being in two different levels of N for these two different traits agrees with results from other studies (Bryant et al., 1986; Wade et al., 1996; Yu and Bernardo, 2004).



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 1. Testcross genetic variance (VTC) pooled for each number of parents.

 
The expected pattern of decline in genetic variance with decreasing N under a purely additive model (shown as solid line with shaded markers in Fig. 1) was closely followed by grain moisture and plant height. Grain moisture VTC was slightly greater at N = 4 than at N = 8. Plant height VTC decreased steadily as the number of parents decreased from N = 8 to N = 1. Despite the relatively strong relationship between VTC and N in these traits, however, using N = 4 or 8 parents instead of only N = 2 did not lead to a significantly higher VTC for both traits.

Grain yield and ear height, on the other hand, had VTC that did not follow the additive pattern as closely as did grain moisture and plant height. For grain yield, the observed VTC (7.09 Mg2 ha–2) at N = 2 was numerically lower than the expected VTC (10.43 Mg2 ha–2). The observed VTC (15.20 Mg2 ha–2) at N = 4 was numerically higher than the expected VTC (12.17 Mg2 ha–2) at N = 4, resulting in a change of rank. The observed VTC at N = 4 was twice that at N = 2, clearly indicating that using four parents resulted in a larger genetic variance compared with using only two (Table 2). Increasing the number of parents further to eight, however, did not result in a higher genetic variance than what was already observed at N = 4. Specifically, VTC at N = 1, 2, and 8 were not significantly different.


View this table:
[in this window]
[in a new window]
  		Table 2. Pooled estimates of testcross genetic variance (VTC) in maize populations with different numbers of parents.

 
For ear height, the observed VTC (18.03 cm2) at N = 2 was numerically higher than the expected VTC (14.02 cm2). This trend was reversed at N = 4, where the observed value (13.60 cm2) was lower than expected (16.36 cm2). Thus, increasing N from 2 to 4 ironically resulted in a 25% decrease in observed VTC. This decrease was not statistically significant, although the VTC at N = 4 was already almost as low as that at N = 1 (Table 2). Increasing N further to eight reversed the reduction, but not to as high as what was attained at N = 2.

Mean
Inbreeding increases with smaller N, so inbreeding depression would be most intense at N = 1. Testcross means, however, behave in an additive manner even if dominance is present (Bernardo, 2002, p. 79). Hence, testcross means were expected to be the same for the four levels of N under an additive genetic model. But the results showed otherwise. Testcross means at N = 1 and 8 tended to differ slightly but significantly from the means at N = 2 and 4 (Table 3). For grain yield and grain moisture, the mean at N = 8 was lower than the mean at other levels of N. For plant height, the mean at N = 1 was lower, whereas the mean at N = 8 was higher, than that at N = 2 and 4. For ear height, the mean tended to be lower at N = 1 and 8 than at N = 2 and 4.


View this table:
[in this window]
[in a new window]
  		Table 3. Testcross means and predicted grain yield of the upper proportion (p) of maize populations with different numbers of parents.

 
The differences observed in testcross means therefore suggest a nonadditive genetic model. Epistasis may cause a nonlinear relationship between the mean and N (Falconer and Mackay, 1996, p. 250). And whereas the two generations of random mating that were performed would have dissipated much of the gametic disequilibrium in each population, the means would still have been affected by any remaining gametic disequilibrium (Melchinger, 1987). In addition, inbreeding may affect different traits differently (Bryant et al., 1986; Wade et al., 1996; Naciri-Graven and Goudet, 2003), depending on the complexity of their genetic architecture (e.g., fitness-related vs. morphologic traits). Hence the mean of one trait may behave differently from the mean of another type of trait when linkage disequilibrium is present. These three factors (i.e., epistasis, gametic disequilibrium, and trait type) may have caused the means to deviate unpredictably as a function of N.

Predicted Mean
There was no change in rank according to selection intensity among all levels of N for predicted yield based on the usefulness criterion (Table 3). Selecting 1% of the population (i.e., p = 0.01, the highest selection intensity) always resulted in the highest predicted mean, whereas selecting 20% (i.e., p = 0.20, the lowest selection intensity) always resulted in the lowest predicted mean. The highest predicted mean (8.64 Mg ha–1) was observed in N = 4 at p = 0.01; the lowest (8.32 Mg ha–1) was observed in N = 8 at p = 0.20. The N = 4 population had the highest predicted mean in all selection intensities. Under the lowest selection intensity (p = 0.20), the predicted mean (8.45 Mg ha–1) at N = 4 was still higher than the lowest mean under p = 0.01 (8.43 Mg ha–1), which was that at N = 2. The overall highest and lowest values indicated, however, that the yield differences among the different levels of N and p were very small.

The importance of genetic variance in contributing to a superior population was demonstrated in N = 8. The mean grain yield was significantly lower at this level of N compared with the mean at N = 2. Yet because of a VTC that was almost as high as that of N = 4 (i.e., the highest VTC), the predicted yield became as high as that of N = 2 under p = 0.10 and 0.20 and even slightly higher under p = 0.01.

Average Heterozygosity
The average number of alleles per locus was 1.3 for N = 1, 1.7 for N = 2, 1.8 for N = 4, and 2.0 for N = 8. The average marker heterozygosity (HM) was 0.34 for N = 1, 0.67 for N = 2, and 0.76 for N = 4. The HM at N = 8 was 1 because at this level, all the loci were heterozygous as only data from polymorphic SSRs were used. As was expected, average marker heterozygosity (HM) decreased with N (Fig. 2) and had a near-linear association (R2 = 0.95) with expected heterozygosity. Average heterozygosity was the first of two terms in the model being tested, i.e., VTC = (1 – 1/2N)VTC(N = {infty}). The quantity 1/2N represents the amount of heterozygosity lost (i.e., inbreeding) because of a finite N. The results therefore confirm the decrease in heterozygosity as the number of parents in the testcross populations decreased. Any nonlinear decrease of genetic variance with inbreeding in these populations was therefore not due to inconsistent levels of heterozygosity as a function of N in the testcross populations.



View larger version (12K):
[in this window]
[in a new window]
  		Fig. 2. Expected heterozygosity (HE) and average SSR marker heterozygosity (HM) for populations with different numbers of parents.

 
Genetic Variance and Number of Parents
Advanced cycle breeding is the overwhelming use and reuse of elite inbreds in a breeding program (Bernardo, 2002, p. 70). Under an additive genetic model, it compromises genetic variation in two ways. First, using elite inbreds from the same heterotic group leads to a higher level of genetic relatedness within breeding programs over time. Second, the predominant use of only two inbred parents limits the total possible number of alleles per locus to only two.

In this study, genetic variance, particularly for yield, did not behave as expected from an additive genetic model. The grain yield VTC at N = 4, which was comparable to VTC(N = {infty}), strongly implied an excess of VA in this subpopulation that was supposed to suffer from a 12.5% loss of genetic variation relative to the base population. Other studies also reported an increase in genetic variance in certain bottleneck sizes and traits (Cheverud and Routman, 1996; López-Fanjul and Villaverde, 1989; Wade et al., 1996). On the other hand, the steep decrease in VTC for grain yield at N = 2 was numerically more than the expected 25% loss of genetic variance. Epistasis apparently did not help alleviate the loss of genetic variance in this particular population size. Genetic variance that was lower than what is predicted by additive gene effects was also observed by Yu and Bernardo (2004). This steep loss of genetic variance seems contradictory to the seemingly lasting supply of genetic variation in U.S. maize breeding programs, even with the predominant use of only two parents (Tracy et al., 2004). It is possible that the original variability present in the original progenitors of advanced cycle inbreds might simply be still undepleted at the current state of U.S. maize breeding programs.

In addition, plant breeders in effect have two levels of variation to choose from in their breeding programs: variance among populations and variance within populations. An increase in the variance among populations would compensate for a decrease in the variance within populations. Our study addresses within-population variation only, but our results nonetheless indicate that within-population variation does not always behave as expected from an additive genetic model.

The higher VTC and predicted mean for grain yield at N = 4 versus that at N = 2 suggest the importance of using multiple parents in creating breeding populations, which could eventually help sustain genetic variability in breeding programs. The use of multiple parents, however, is not a common practice among maize breeders. In addition, this study tried to capture as much as possible the full amount of variability that could arise from the use of multiple parents by evaluating a large number of families (i.e., 200–250 per level of N). A breeding population of this size is probably larger than what most breeders manage in practice.

Although epistasis is often difficult to detect (Hinze and Lamkey, 2003), some reports have shown significant epistatic effects for maize yield in certain hybrid combinations (Lamkey et al., 1995; Wolf and Hallauer, 1997; Eta-Ndu and Openshaw, 1999). In a recurrent selection program that recombined only five lines, Guzman and Lamkey (2000) found no significant decrease in additive genetic variance for maize grain yield after five generations. They explained that epistasis was the most probable reason for the lack of a short-term loss in genetic variation. At the molecular level, nonallelic interactions have similarly been documented in crosses between common cultivars in other species such as rice (Li et al., 1997), soybean [Glycine max (L.) Merr., Lark et al., 1995], and common bean (Phaseolus vulgaris L., Johnson and Gepts, 2002). Our study therefore supports what other studies have suggested regarding the role of epistasis in elite germplasm.


   ACKNOWLEDGMENTS
 
We acknowledge support for the completion of this study from the Troyer/Darwin Fellowship and Pioneer Hi-Bred International.

Received for publication January 3, 2005.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 




This article has been cited by other articles:


Home page
 Crop Sci.Home page
V. M. Rodriguez, A. Butron, G. Sandoya, A. Ordas, and P. Revilla
Combining Maize Base Germplasm for Cold Tolerance Breeding
Crop Sci., July 30, 2007; 47(4): 1467 - 1474.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Tabanao, D. A.
Right arrow Articles by Bernardo, R.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Tabanao, D. A.
Right arrow Articles by Bernardo, R.
Agricola
Right arrow Articles by Tabanao, D. A.
Right arrow Articles by Bernardo, R.
Related Collections
Right arrow Crop Genetics
Right arrow Crop Models

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Vadose Zone Journal
Journal of Plant Registrations Soil Science Society of America Journal
Journal of Natural Resources
and Life Sciences Education		 Journal of
Environmental Quality