Skip to main content
NCBI home page
Genetics logoLink to Genetics
. 2006 Jan;172(1):457–465. doi: 10.1534/genetics.105.040899

Genetic Variation and Selection Response in Model Breeding Populations of Brassica rapa Following a Diversity Bottleneck

William H Briggs 1,1, Irwin L Goldman 1
PMCID: PMC1456173  PMID: 16157675

Abstract

Domestication and breeding share a common feature of population bottlenecks followed by significant genetic gain. To date, no crop models for investigating the evolution of genetic variance, selection response, and population diversity following bottlenecks have been developed. We developed a model artificial selection system in the laboratory using rapid-cycling Brassica rapa. Responses to 10 cycles of recurrent selection for cotyledon size were compared across a broad population founded with 200 individuals, three bottleneck populations initiated with two individuals each, and unselected controls. Additive genetic variance and heritability were significantly larger in the bottleneck populations prior to selection and this corresponded to a heightened response of bottleneck populations during the first three cycles. However, the overall response was ultimately greater and more sustained in the broad population. AFLP marker analyses revealed the pattern and extent of population subdivision were unaffected by a bottleneck even though the diversity retained in a selection population was significantly limited. Rapid gain in genetically more uniform bottlenecked populations, particularly in the short term, may offer an explanation for why domesticators and breeders have realized significant selection progress over relatively short time periods.

POPULATION bottlenecks are prolific in domesticated plant and animal populations historically and in modern breeding practices. Restrictions in population size have recently been a focus of numerous studies in evolutionary genetics regarding population subdivision and inbreeding (Saccheri et al. 2001), founder effects (Goodnight 1988), and domestication (Whitt et al. 2002). Following a bottleneck, additive genetic variance (VA) may increase due to contributions from dominance (Robertson 1952; Willis and Orr 1993) and epistatic variances (Cheverud and Routman 1996; Naciri-Graven and Goudet 2003), in particular for traits closely related to fitness (Zhang et al. 2004). This has special relevance to breeders, as the trait of interest is inextricably tied to fitness through selection. The increases in VA observed for some traits following a restriction in population size are in contrast to a purely additive model, which predicts a decline in VA parallel to the loss of genetic diversity caused by the bottleneck (Nei et al. 1975). Empirical studies in various laboratory animal species have confirmed that VA for a quantitative trait can increase following a bottleneck (Bryant et al. 1986; M; Fernandez et al. 1995; Ruano et al. 1996; Cheverud et al. 1999). Under most circumstances, VA is the only part of the genetic variance that is heritable and contributing to selection response.

An increase in VA may enhance selection response. This potential gain from inbreeding has strong implications for agricultural improvement programs, particularly when a rapid response to selection and phenotypic uniformity are desired. Increases in VA following population subdivision in maize suggest that random inbreeding may have desirable effects and that too much diversity in a population may dampen selection response (Edwards and Lamkey 2003). The strong effect of bottlenecks on genetic variance merits further investigation into their influence on important breeding factors such as genetic variance distribution, selection response rate and duration, and population diversity and divergence.

Recurrent selection populations are a useful system for studying population and quantitative genetic parameters relevant to agricultural improvement. Recurrent selection is a cyclical process for advancing a population for one or more traits. With each cycle, parents that are superior are selected and intermated to produce progeny for the next round of selection. The expectation is that favorable alleles at loci will increase in frequency and assemble together through recombination with advancing generations. Recurrent selection is the predominant method of plant and animal breeders for developing improved populations of domesticated species. Evolutionary biologists have also extensively used recurrent selection populations in model animal species, referred to as artificial selection experiments, for developing many fundamental concepts in quantitative inheritance (Falconer 1992; Hill and Caballero 1992). However, recurrent selection does not lend itself well to experimentation because it can take many years and even decades to reveal substantial phenotypic changes from selection in field experiments, even in annual plant species. Additionally, it is difficult to incorporate experimental controls in such experiments because of resource limitations. To date, no laboratory-based crop models have been developed to investigate selection response and population diversity following a restriction in population size.

We developed a model artificial selection system in the laboratory using rapid-cycling Brassica rapa to investigate aspects of diversity and selection response following a population bottleneck. Rapid-cycling B. rapa, developed by Williams and Hill (1986), was suitable because it is outcrossing, diminutive, grows under fluorescent lighting in a laboratory, and has a generation time of 36 days. An advantage of plants as an artificial selection system is the ease with which progeny can be stored for years as seed for simultaneous evaluation of multiple generations, subsequent crosses, and molecular genetic analyses. In this study, four recurrent selection populations were constructed to model the effects of a bottleneck on genetic variance, response to recurrent selection, and population diversity and structure. One population was founded with 200 random individuals. The other three populations were each initiated with two random individuals, a restriction typical of a breeding program.

MATERIALS AND METHODS

Population construction and recurrent selection:

Four sets of recurrent selection populations were established (Figure 1) by mating random individuals from the rapid-cycling B. rapa population 1-1 (Crucifer Genetics Cooperative, Madison, WI). One base population was started by mass pollinating 200 plants. Pollinations were made using wands constructed of lyophilized bee thoraxes (Carolina Biological Supply, Burlington, NC) fastened to toothpicks. The progeny from these plants were the starting material for a broad recurrent selection population. A composite of equal numbers of seed from each of the 200 founders was made. The other three populations, bottleneck 1, 2, and 3, were each founded by a cross of two random plants. The progeny from each cross were grown and massed in isolation from each other. Seeds from these masses were used as the base for each bottleneck recurrent selection population. An equal number of seeds from each plant was used to create a base composite.

Figure 1.

Figure 1.

Open in a new tab

Schematic of the artificial selection experiment. One broad and three bottleneck populations were founded with random individuals taken from the rapid-cycling B. rapa stock. These population bases were each subdivided into two recurrent selection subpopulations divergent for cotyledon size and one random control subpopulation.

Divergent biparental mass selection was initiated using the same procedure for each of the four base populations. For the first cycle, three flats of 200 plants were grown and evaluated visually for cotyledon size on the seventh day after sowing. The 25 individuals with the largest cotyledons (12.5%) were selected from the first flat of plants, while the 25 with the smallest cotyledons were taken from the second flat to initiate large and small cotyledon subpopulations, respectively. A random subset of 25 plants was taken from the third flat of 200 plants to establish a control subpopulation. In total, four sets of three subpopulations were established, each set consisting of a large, small, and control subpopulation derived from a single broad or bottleneck base population. Each of the 12 was housed within a porous pollination bag (Vilutis Inc., Frankfort, IL) to prevent cross-pollination between populations. Pollinations were made once daily for 10 days, beginning the day after the first flower opened in a subpopulation. Pollen was collected from all open flowers onto the pollination wand and then applied to all exposed stigmas. Seed was harvested from each plant individually. An equal number of seeds from every plant was mixed to form a composite for each subpopulation. This composite of progeny was used to grow the next cycle of plants for selection. The remaining seed was stored for simultaneous evaluation of all of the selection cycles.

For all subsequent cycles of selection, 200 progeny from the previous cycle were grown and evaluated for cotyledon size on day 7. The 25 plants with the largest or smallest cotyledons were selected and massed for the large and small subpopulations, respectively. Twenty-five individuals were taken at random and massed to maintain the control subpopulations. During all selection cycles, plants without meristematic growth were not selected, as these plants have characteristically large cotyledons and abnormal development. Seed was harvested and stored as described. Ten cycles of selection were completed within each subpopulation.

Estimation of genetic variances:

A North Carolina design I mating scheme (Comstock and Robinson 1948) was used to estimate genetic variance components for cotyledon size in each base population prior to selection. This is a two-factor nested design assuming no epistatic variance. For each population, 50 random individuals were designated as males. Each male was crossed as follows to four random females, for a total of 200 crosses per population. Any open flowers were removed from the designated females. Four to six completely closed buds were opened and the anthers removed. Pollen was collected from the designated pollinator (male) onto a beestick and applied to the stigma of the emasculated buds on the female plant. All flowers subsequent to pollination were removed. The progeny from each cross was a full-sib family, while the progenies of the four crosses of a male parent were considered collectively as a half-sib family.

Twenty progeny from each full-sib family were planted in a completely randomized design with two replications to estimate genetic variance components and heritability for cotyledon size in each of the base populations. Progeny were grown one seed per cell in rows of 20 cells. Seven days after sowing, one cotyledon was pulled from each plant and pressed. Cotyledons were scanned at a resolution of 150 dpi on an Epson scanner. The surface area of each cotyledon was determined using ImageJ public domain software v. 1.28 (National Institutes of Health 2002).

A linear mixed effects model analysis of variance for cotyledon size was performed for each population's mating design using SAS statistical software (SAS Institute, Cary, NC). All effects, replications, males, and females nested within males were considered random. The mivque0 method of the varcomp procedure was used to estimate variance components. This method produces quadratic unbiased estimates of variance. Additive genetic variance (VA) was estimated as four times the variance among males. Dominance genetic variance (VD) was estimated as four times the difference between the variance among females within males and the variance among males. Estimates of the phenotypic variance (VP) were calculated as the sum of the VA, VD, and residual error variance. Narrow-sense heritability (h2) was the ratio of VA to VP. One thousand iterations of parametric bootstrapping with 50% subsampling were used to define 95% confidence intervals for each parameter estimate (Reverter et al. 1998).

Evaluation of selection:

Remnant seeds from rapid-cycling B. rapa population 1-1, the broad and bottleneck base populations, and every cycle of all 12 recurrent selection subpopulations were grown in a completely randomized design with three replicates. The replicates were initiated over 3-week intervals. A row of 10 seeds from each entry was sown per replicate. Seven days after sowing, one cotyledon was removed from each plant and scanned at 150 dpi. Cotyledon surface area was determined using ImageJ. Cycle means were estimated by maximum likelihood for each population, using a mixed variable model with replicate as a random effect, subpopulation (direction of selection), and cycle as fixed effects. Response per cycle was reported as the slope coefficient of a linear regression of each trait vs. cycle. Response to selection for cotyledon size was compared among the four populations and within each set of subpopulations by multiple regression. The full model for testing contrasts was determined by step-wise addition, using Akaike's information criteria (AIC) to add significant parameters to a base model having separate slopes and intercepts for all 12 subpopulations.

Analysis of genetic diversity and population structure:

Plants were sampled from the broad and the bottleneck 1 base populations before selection, as well as their derivative large, small, and control subpopulations following 10 cycles of selection. A sample of 47 plants was grown from each subpopulation. DNA was isolated using a CTAB extraction from young leaf tissue as described (Futterer et al. 1995).

Restriction digests of genomic DNA with EcoRI and MseI, followed by adaptor ligations and preselective PCR were performed as described (Vos et al. 1995). Ten primer pair combinations were tested on a set of 12 individuals from the bottleneck population at cycle 0. Four primer pair combinations having 10 or more polymorphic bands with consistent size estimation across three independent runs of the test set were chosen for the analysis (GACTGCGTACCAATTCACG + GATCAGTCCTGAGTAACAG, GACTGCGTACCAATTCAGG + GATCAGTCCTGAGTAACAG, GACTGCGTACCAATTCACC + GATCAGTCCTGAGTAACAC, and GACTGCGTACCAATTCACC + GATCAGTCCTGAGTAACTC). Additional polymorphic fragments amplified with these four primer pairs were detected and scored upon analysis of the entire set of samples. Selective PCR primers for the EcoRI-cleaved side of the fragments were labeled at the 5′-terminus with the fluorophore 6-FAM. Final PCR fragments were purified with CleanSeq magnetic beads (Agencourt, Beverly, MA) and mixed with GeneFlo 625 Rox-labeled internal size standards (Chimerx, Milwaukee). PCR fragments were separated by capillary electrophoresis on an ABI 3100 automated sequencer (Applied Biosystems). The sizes of PCR fragments ranging from 50 to 625 bp were estimated using GeneScan software, version 3.1 (Perkin-Elmer). Only fragments that could be reproduced within a 1-bp range were scored. Polymorphic fragments were scored as either present (1) or absent (0) for all sampled plants.

Genetic dissimilarity between each pair of sampled plants was calculated using Jaccard's distance coefficient (Jaccard 1908). This measurement takes into account only 1-1 matches. These matches are more informative than 0-0 matches, as the failure of a band to amplify could occur for a number of reasons and may not reflect identity by descent. The diversity within and the divergence between subpopulations were visualized with multidimensional scale (MDS) plots based on the genetic dissimilarity matrix of Jaccard's coefficients. Genetic diversity and population structure statistics were calculated with the program AFLP-SURV 1.0 (Vekemans 2002) using all marker loci except those that were monomorphic across all of the populations. This program uses the approach of Lynch and Milligan (1994) to calculate population genetic parameters on the basis of the expected heterozygosity of dominant marker loci. The genetic diversity was compared between subpopulations using Nei's gene diversity or expected heterozygosity (HJ). The divergence between the base populations and their derivative selection subpopulations were evaluated using pairwise FST values. Bootstrap confidence intervals were calculated for the FST values by performing 10,000 iterations of sampling with replacement.

The effective population size (Ne) of the broad and bottleneck 1 subpopulations was estimated from temporal changes in the inferred AFLP allele frequencies from cycle 0 to cycle 10. The maximum likelihood estimates of Ne were determined by a coalescent-based model developed by Berthier et al. (2002). Markov chain Monte Carlo was used to calculate the estimates and 95% confidence intervals for Ne using the software CoNe (Anderson 2005).

RESULTS

Genetic variance prior to selection:

Response to selection is directly dependent upon the amount and proportion of genetic variance that is additive. Following a bottleneck, VA has increased for various quantitative traits related to fitness, but decreased in traits with simpler inheritance in several animal experiments (Saccheri et al. 2001). No direct evaluations of this kind have been made in plant breeding populations. We constructed a North Carolina design I mating scheme to estimate the genetic variance components for cotyledon size of each of the model broad and bottleneck populations prior to subdivision and recurrent selection. All F-tests for replication of progeny evaluation were insignificant. Cotyledon size was normally distributed for each of the populations. The initial population means of bottlenecks 1, 2, and 3, were significantly lower than the mean of the broad population (Table 1).The average reduction in cotyledon surface area due to the bottleneck was 0.763 cm2 or 42%. There were no significant differences among the bottleneck means.

TABLE 1.

Population statistics prior to selection

Population Mean cotyledon size (cm2) Heterozygosity VA VD VP h2
Broad 1.82 ± 0.08 0.341 ± 0.016 0.021 (0.016, 0.024) 0.269 (0.264, 0.281) 0.511 (0.508, 0.519) 0.044 (0.034, 0.053)
Bottleneck 1 1.02 ± 0.10 0.262 ± 0.019 0.054 (0.049, 0.059) 0.039 (0.029, 0.049) 0.325 (0.319, 0.331) 0.180 (0.160, 0.200)
Bottleneck 2 1.01 ± 0.15 — 0.029 (0.027, 0.030) 0.023 (0.019, 0.026) 0.152 (0.149, 0.154) 0.198 (0.182, 0.211)
Bottleneck 3 1.14 ± 0.13 — 0.110 (0.102, 0.113) −0.003 (−0.010, 0.012) 0.317 (0.313, 0.326) 0.370 (0.332, 0.383)
Open in a new tab

Each estimate is followed by a standard error or a 95% confidence interval.

VA for cotyledon size was larger in all of the bottleneck populations than in the broad population. The three bottleneck populations all differed significantly from each other for VA, with bottlenecks 1 and 3 having substantially larger estimates than bottleneck 2. The estimated VD for the broad population was nearly 13 times larger than the respective VA. VD was much smaller in the bottleneck populations, with values smaller than the respective VA. The estimate for VD in bottleneck 3 was negative but not significantly different from zero. VP was significantly larger in the broad population compared to all of the bottlenecks. The narrow-sense heritability (h2) was at least threefold larger for all bottleneck populations compared to the broad population. This was particularly apparent in bottleneck 3, which was approximately 10-fold higher. All three bottlenecks had larger h2 due to elevated VA and smaller VD and error variances.

The increase in VA following a bottleneck does not agree with an additive model of inheritance for cotyledon size. Under pure additivity, VA will decrease proportionally with the average degree of heterozygosity in a population (Falconer and Mackay 1996). In a finite population, the average per locus heterozygosity is reduced by 1/2N (Nei et al. 1975). Thus, the average heterozygosity of a population that has been through a bottleneck of two individuals should decline by a quarter. Estimates of the average expected heterozygosity within the broad and bottleneck 1 populations based on AFLP data indicate this approximate reduction in genetic variation (Table 1). The ratio of heterozygosity between the bottleneck 1 and broad populations is 0.77.

Selection response:

The increases in VA observed in the bottleneck populations are suggestive a boost in response to selection for cotyledon size. We carried out 10 cycles of divergent recurrent selection in all of the populations to observe differences in short- and long-term response patterns. Questions of particular interest are how quickly a significant response to selection is achieved, how long selection responses are sustained before attenuation is reached, and how large are the overall responses and response rates in a broad vs. a bottlenecked population.

Cycle means within populations were compared by protected LSD tests. This enabled us to determine how many cycles of selection were performed until a significant response in selection was achieved. The large and small subpopulations of the broad and bottleneck populations all diverged significantly for cotyledon surface area following 10 cycles of selection (Figure 2). The subpopulation divergence was significantly greater in the broad population than in the bottleneck populations (Table 2). Bottleneck 1 had a greater selection response than the other bottleneck populations. The broad population and bottlenecks 1 and 3 had diverged significantly by selection cycle 2. Bottleneck 2 did not have a significant subpopulation divergence until cycle 4. Response to selection in bottlenecks 1 and 3 surpassed the broad population over the first three cycles. Beyond cycle 5, response was no longer appreciable in any of the bottlenecks while a steady response continued in the broad population.

Figure 2.

Figure 2.

Open in a new tab

Cotyledon surface area over 10 cycles of divergent biparental recurrent mass selection for cotyledon size in one broad (200 founders) and three bottleneck (2 founders each) populations of rapid-cycling B. rapa (graphs). Scanned cotyledons from selection cycle 10 progeny (right).

TABLE 2.

Selection response for cotyledon size

Divergence (cm2)a
Response rate (cm2/cycle)b
Population Cycle 1 Cycle 4 Cycle 7 Cycle 10 Large Control Small
Broad 0.237 ± 0.124 0.734 ± 0.123 1.217 ± 0.143 1.251 ± 0.118 0.052 −0.020 −0.056
Bottleneck 1 0.238 ± 0.130 0.705 ± 0.125 0.445 ± 0.172 0.837 ± 0.124 0.025 — −0.024
Bottleneck 2 0.199 ± 0.133 0.327 ± 0.131 0.666 ± 0.131 0.350 ± 0.122 0.041 0.034 —
Bottleneck 3 0.231 ± 0.132 0.721 ± 0.123 0.892 ± 0.176 0.582 ± 0.121 0.037 — —
Open in a new tab
a

Divergence was calculated as the difference in large and small subpopulation means.

b

Only significant regression coefficients (P < 0.05) are reported.

We used multiple regression to analyze selection response rates. This approach allowed for tests of selection response significance within each subpopulation and comparisons of response across populations using contrasts. Response per selection cycle for cotyledon size was greatest in the broad population. A contrast with the three bottleneck populations indicated a difference in response rate of 0.0252 cm2/cycle (P < 0.05). Regression coefficients for downward selection were significant only for the broad and bottleneck 1 populations. Selection for small cotyledons was substantially more successful in the broad population. A decrease in cotyledon size was noted in the broad control but this was a significantly smaller response than that of the parallel small subpopulation. An increase in cotyledon size was found in the bottleneck 2 control. This was not found to differ from the response per cycle in the corresponding large subpopulation, indicating a significant rise in cotyledon size in this control.

Population diversity and structure:

The degree of uniformity and the amount of genetic variation retained in a population following a bottleneck and selection is unclear. The uniformity of a population for a trait of interest is not necessarily reflective of the average level of gene diversity or the divergence of a population from its source. To understand the pattern of diversity retained in the model populations, we sampled individuals from the broad and bottleneck 1 before and after 10 cycles of recurrent selection.

A total of 83 AFLP fragments were scored for each individual sampled. Fifteen of these were present in all of the plants evaluated; however, it was not possible to distinguish between individuals that were heterozygous and homozygous for the presence of a fragment. The broad population base prior to selection was polymorphic for the other 68 fragments. A subset of 53 fragments was polymorphic in the bottleneck population (Table 3).None of the fragments were polymorphic exclusively in the bottleneck. The number of polymorphic fragments decreased in all of the subpopulations following 10 cycles of selection. The number of fragments that reached fixation in the small and control subpopulations was substantially greater than in the large subpopulations in both the broad and bottleneck 1 cases.

TABLE 3.

Population diversity and divergence statistics before and after selection

Population Subpopulation No. of polymorphic loci Nei's gene diversity (HJ ± SE) Pairwise FST, divergence from base (95% CI) Ne based on temporal changes in marker frequencies (95% CI)
Broad Base (cycle 0) 68 0.341 ± 0.016 — —
Large (cycle 10) 59 0.309 ± 0.018 0.087 (0.051, 0.125) 19.2 (15.2, 24.5)
Control (cycle 10) 41 0.225 ± 0.019 0.201 (0.145, 0.254) 9.4 (7.8, 11.8)
Small (cycle 10) 48 0.228 ± 0.020 0.171 (0.101, 0.246) 11.0 (9.3, 13.2)
Bottleneck 1 Base (cycle 0) 53 0.262 ± 0.019 — —
Large (cycle 10) 41 0.215 ± 0.021 0.119 (0.067, 0.167) 14.0 (11.2, 17.4)
Control (cycle 10) 23 0.132 ± 0.020 0.333 (0.197, 0.369) 7.2 (6.5, 7.6)
Small (cycle 10) 26 0.136 ± 0.019 0.320 (0.182, 0.361) 7.7 (6.7, 7.6)
Open in a new tab

We used the AFLP data to estimate the diversity before and after selection to evaluate and compare the extent of change in the broad and bottleneck populations from selection cycles 0 to 10. Although dominant markers do not provide a direct measure of heterozygosity, we were able to infer the frequency of heterozygotes by assuming Hardy-Weinberg equilibrium within our samples. Under these conditions, the frequency of heterozygotes would be 2(q − q2), where q is the frequency of marker absence. This enabled the calculation of Nei's gene diversity (HJ). Diversity decreased in all of the subpopulations during 10 cycles of recurrent selection (Table 3). This decrease was particularly drastic in the control and small subpopulations. The proportion of diversity lost from cycle 0 to cycle 10 was always more severe in the bottleneck subpopulations compared with their broad counterparts. The level of diversity before and after selection was significantly smaller in the bottleneck samples.

The patterns of subpopulation divergence from the broad and bottleneck base populations are apparent in the three-dimensional MDS plots (Figure 3). The plots can be compared directly as they were generated from a single dissimilarity matrix. Superimposition of the broad and bottleneck MDS plots reveals that the two populations are also diverged from each other. We evaluated the integrity of population substructuring by calculating pairwise FST values between each of the subpopulations at cycle 10 and their respective base population before selection (Table 3). Bootstrap confidence intervals allowed us to compare the genetic divergence within and between populations. After 10 cycles of selection, all of the subpopulations, including the controls, had diverged from the base populations. The large subpopulations diverged from their population base significantly less than the control and small subpopulations in both the broad and bottleneck cases. The amount and pattern of subpopulation divergence did not differ between the populations.

Figure 3.

Figure 3.

Open in a new tab

MDS plots of individuals sampled from the broad and bottleneck populations before (base) and after (large, control, and small) 10 cycles of selection. The MDS positions for both plots were determined using a single matrix of genetic distance estimates based on AFLP similarities. All axes have a tick distance of 0.1 MDS units.

To compare the effective population size of the subpopulations on the basis of genetic estimates and demographic data, we calculated Ne from temporal changes in allele frequencies inferred from the AFLP data (Table 3). According to the selection scheme, the demographic Ne is ideally 25. All of the estimates of Ne based on genetic data were significantly <25. It was infrequent for all 25 plants selected from any subpopulation to produce seed. Nonetheless, there were never <13 female parents per cycle. By contrast, the genetic Ne was significantly <13 for both the broad control and bottleneck 1 control and small subpopulations.

DISCUSSION

Model system:

Quantitative theory in plant breeding has generally been supported by retrospectively analyzing data derived from populations used for crop improvement that lack controls. A number of fundamental questions in plant breeding could be addressed using a model system that lends itself more favorably to experimental work than established crop systems. Maize, the traditional species for such analyses, cannot be grown in substantial numbers in a laboratory or even a greenhouse setting. Furthermore, only three generations can be accomplished in 1 year at best. The life cycle of most other crops is even longer than that of maize. In comparison, 10 generations can be achieved per year in rapid-cycling B. rapa.

We used B. rapa and the trait of cotyledon size as a model system for breeding experiments. A comparison of selection responses in our populations with applied breeding and artificial selection populations indicates that the B. rapa model is representative of recurrent selection experiments. The responses to upward and downward selection for cotyledon size expressed as a ratio of cycle 0 to cycle 10 means for each of the populations spanned from 1.03 for the bottleneck 2 small subpopulation to 1.83 for the broad large subpopulation. These fall close to the responses of several recurrent selection experiments in maize, poultry, and laboratory mice of varying population sizes and selection intensities. In maize, cycle 0 to 10 ratios for high oil and protein (Dudley and Lambert 2004), prolificacy (Maita and Coors 1996), ear length (Cortez-Mendoza and Hallauer 1979), and others (Hallauer and Miranda 1988) ranged from 1.0 to 2.0. Response values were similar for chickens, turkeys, and mice (Hill and Bunger 2004).

Population bottlenecks and selection potential:

Population bottlenecks of two individuals caused an increase in VA for cotyledon surface area in three populations of rapid-cycling B. rapa, relative to a broadly based population. This effect of restricted population size is in agreement with theoretical predictions of quantitative traits, which include dominance and epistatic effects (Goodnight 1988; Willis and Orr 1993; Edwards and Lamkey 2003). Several studies of quantitative traits in animals have had similar results (Bryant et al. 1986; Lopez-Fanjul and Villaverde 1989; Fernandez et al. 1995; Ruano et al. 1996; Cheverud et al. 1999). These studies have reported values only for VA and VP by parent–offspring regression, but not VD. The nested mating design used in our study enables derivation of VD. The VD estimates for the bottleneck are much smaller than those for for the broad population. If an increase in VD occurs due to increases in rare recessives, the VD associated with these genes is not expected to change appreciably. The degree of dominance in the broad population is atypically large compared with VD/VA ratios for several traits in maize and other species (Hallauer and Miranda 1988). Interlocus interactions may also be important in the genetics controlling cotyledon size; however, the mating design does not evaluate epistasis. Bottleneck 3 had a negative estimate of VD. The probability of a negative estimate of VD in a North Carolina design I was >0.25 across a range of dominance variance values in simulation studies (Bridges and Knapp 1987). As VD for the other bottlenecks are small, the probability of a negative estimate may be quite high.

The increase in heritability accompanying a population bottleneck leads to the prediction of a greater immediate selection response. An increase in selection response following a bottleneck has occurred in studies of inbred populations of houseflies (Bryant and Meffert 1995) and Drosophila melanogaster (Lopez-Fanjul and Villaverde 1989). Rapid gain in bottlenecked populations, particularly in the short term, may offer an explanation for why both crop domesticators and breeders have realized significant selection progress over relatively short time periods. In particular, the initial stages of the inbred–hybrid method of breeding are characterized by subdivision of populations into inbred lines using severe inbreeding or self-pollination. This method has largely dominated plant breeding in many parts of the world during the 20th century. Practitioners of this method may have been rewarded with early selection gains following close inbreeding, thereby contributing to the widespread adoption of these types of breeding strategies.

The initial boost in heritability did not lead to larger phenotypic means than non-inbred lines because of inbreeding depression. There appeared to be inbreeding depression for cotyledon size in this study, although cotyledon size may have been smaller in the three bottlenecks due simply to drift. The plants used to establish the bottleneck populations were taken at random. If they had also been selected for cotyledon size, an increase in VA may have occurred without inbreeding depression, particularly if the increased size were due to favorable alleles at different loci in the two individuals. Cotyledon size is likely genetically complex and pleiotropic with other traits affecting fitness. Had a quantitative trait with less effect on fitness been chosen, it is conceivable that the apparent inbreeding depression would not have accompanied an increase in VA.

Estimation of the initial genetic variance components in the populations enabled the prediction of selection response. Assuming a population size of 25 and that 200 plants were evaluated, the intensity of selection determined from the truncated normal distribution, i, is 1.636. Predicted response, R, can be calculated as R = ih2σP, where h2 is the narrow-sense heritability and σP is the phenotypic standard deviation (Falconer and Mackay 1996). Using estimates of h2 and σP from our experiment, the predicted response of the first cycle of selection is 0.052 cm2 for the broad population and 0.168, 0.126, and 0.341 cm2 for the bottleneck populations 1, 2, and 3, respectively. The predicted response for the broad population is nearly equivalent to the response per cycle estimated by regression. The response per cycle of the bottleneck populations is much less than the predicted response; however, the response during the first three cycles of selection is much greater than in later cycles.

Dominance and epistatic variance may be more salient for selection response following a population bottleneck, but they can become accessible in a broader recurrent selection population if selection is continued to more advanced cycles (Cockerham and Tachida 1988). The conversion of epistatic into additive variance is dependent not on the severity of a bottleneck but on the level of inbreeding of a population (Naciri-Graven and Goudet 2003). Inbreeding increases with cycles of recurrent selection because of assortative mating regardless of population size. The results of this study imply that a population bottleneck prior to selection may not adversely affect immediate response. In effect, breeders may choose between a rapid but limited response in a narrow genetically more uniform population and a larger more long-term gain in a more diverse population, if our data are reflective of crop populations.

The comparison of broad and bottleneck populations is particularly relevant to allogamous domesticated plant and animal species. However, many important crops are autogamous. Such crops are maintained as homozygous lines through selfing. New varieties are continuously generated by successive rounds of crossing between existing lines and selection among the resulting progeny lines. Recurrent selection is not generally applied to these crops. Perhaps a greater response is achieved in selfing species by making further narrow crosses rather than continuing with additional cycles of selection, as the level of diversity available for further selection in primarily autogamous populations is already low. The practices of plant breeders and the data from this study showing early response and response attenuation in bottleneck populations support such a conclusion.

Controls:

The unselected control within each of the populations is an indicator of the potential magnitude of drift and natural selection that may increase or decrease phenotypic measurements in any of the subpopulations. Most recurrent selection programs do not include a control population. The controls in this study illustrate the extent to which crop improvement programs may inadvertently benefit or suffer because of drift and natural selection.

Both the broad and bottleneck 2 controls exhibited directional changes in cotyledon size across cycles. With the exception of bottleneck 2, directional changes in the controls across cycles were significantly less when contrasted with the large and small selections by regression analysis. However, the drastic reduction in diversity, divergence from the base population, and small genetic Ne that accompany the significant decrease in cotyledon size in the broad control strongly suggest that natural selection was active. The pronounced deviation of Ne in the control and small groups of both the broad and bottleneck 1 populations from the intended population size of 25 was greater than expected. Regardless of the selection pressure imposed by the protocol, cotyledon size is related to fitness. It is conceivable that natural selection for cotyledon size would be effective in the populations, particularly in the bottleneck populations, where purging of deleterious alleles affecting fitness may be occurring. A composite of equal numbers of seed from each selected parent was made for each generation to limit drift and natural selection. However, differences in germination rate and viability among sibships, particularly in the small subpopulations, would have increased inbreeding and decreased Ne on the basis of temporal changes in gene frequencies.

Diversity and the genetic environment during selection:

Changes in the genetic environment associated with a population bottleneck could constitute a genetic revolution according to Ernst Mayr (1954). In his description of the founder effect, any given gene is strongly influenced by its genetic background environment, as it contributes jointly with the action of other genes to a given character. Consequently, in an extreme case an allele that displays a high selective advantage in one background may be selected against in a different genetic environment. In a large well-established variable population, the favorable alleles will be those that produce a viable combination with the greatest number of different genetic backgrounds. However, if a few individuals are isolated from this broad gene pool and found a new population, the relative selective value of an allele may be drastically different as the number of possible genetic interactions is reduced and homozygotes may be exposed. In this condition of decreased heterozygosity and allele number due to genetic drift, alleles viable in the homozygous condition may suddenly have a selective advantage and impact the rate and magnitude of the genetic revolution (Mayr 1963).

Mayr described the combination of drift and the change of selective values of alleles that accompany a population bottleneck as a basis for speciation in nature. Similarly, Wright proposed in his shifting-balance theory that an allele may be favored by selection in one deme with one set of interactions, but selected against in a deme with a distinct genetic background, even in the same environment (Wright 1968). According to Mayr, the single inseminated female exemplifies the founder effect (Mayr 1963). Likewise, a single cross between two individuals is a typical bottleneck in an applied breeding program. In this study we show that a population bottleneck altars the genetic environment in ways that can impact selection by limiting the overall genetic variation available, as measured with the AFLP data, while increasing VA for some traits. Additional investigations into the genetic environment following a bottleneck are crucial for developing a framework for investigating crop domestication and artificial selection.

Acknowledgments

We are grateful for the technical assistance provided by K. Quast, E. Quast, and J. Baeten. B. Larget and A-C. Thuillet provided statistical guidance and useful suggestions. This work was supported in part by a Thomsen Wisconsin Distinguished Graduate Fellowship awarded to W.B.

References

  1. Anderson, E. C., 2005. An efficient Monte Carlo method for estimating Ne from temporally spaced samples using a coalescent-based likelihood. Genetics 170: 955–967. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Berthier, P., M. A. Beaumont, J. M. Cornuet and G. Luikart, 2002. Likelihood-based estimation of the effective populations size using temporal changes in allele frequencies: a genealogical approach. Genetics 160: 741–751. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bridges, W. C., and S. J. Knapp, 1987. Probabilities of negative estimates of genetic variances. Theor. App. Gen. 74: 269–274. [DOI] [PubMed] [Google Scholar]
  4. Bryant, E. H., and L. M. Meffert, 1995. An analysis of selection response in relation to a population bottleneck. Evolution 49: 626–634. [DOI] [PubMed] [Google Scholar]
  5. Bryant, E. H., S. A. McCommas and L. M. Combs, 1986. The effect of an experimental bottleneck upon quantitative genetic variation in the housefly. Genetics 114: 1191–1211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cheverud, J. M., and E. J. Routman, 1996. Epistasis as a source of increased additive genetic variance at population bottlenecks. Evolution 50: 1042–1051. [DOI] [PubMed] [Google Scholar]
  7. Cheverud, J. M., T. T. Vaughn, S. Pletscher, K. King-Ellison, J. Bailiff, et al., 1999. Epistasis and the evolution of additive genetic variance in populations that pass through a bottleneck. Evolution 53: 1009–1018. [DOI] [PubMed] [Google Scholar]
  8. Cockerham, C. C., and H. Tachida, 1988. Permanancy of response to selection for quantitative characters in finite populations. Proc. Natl. Acad. Sci. USA 85: 1563–1565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Comstock, R. E., and H. F. Robinson, 1948. The components of genetic variance in populations of biparental progenies and their use in estimating the average degree of dominance. Biometrics 4: 254–266. [PubMed] [Google Scholar]
  10. Cortez-Mendoza, H., and A. R. Hallauer, 1979. Divergent mass selection for ear length in maize. Crop Sci. 19: 175–178. [Google Scholar]
  11. Dudley, J. W., and R. J. Lambert, 2004. 100 generations of selection for oil and protein in corn. Plant Breed. Rev. 24(I): 79–110. [Google Scholar]
  12. Edwards, J. W., and K. R. Lamkey, 2003. Dominance and genetic drift: predicted effects of population subdivision in a maize population. Crop Sci. 43: 2006–2017. [Google Scholar]
  13. Falconer, D. S., 1992. Early selection experiments. Annu. Rev. Genet. 26: 1–14. [DOI] [PubMed] [Google Scholar]
  14. Falconer, D. S., and T. F. C. Mackay, 1996. Introduction to Quanitative Genetics. Longman Group, Essex, UK.
  15. Fernandez, A., M. A. Toro and C. Lopez-Fanjul, 1995. The effect of inbreeding on the redistribution of genetic variance of fecundity and viability in Tribolium castaneum. Heredity 75: 376–381. [Google Scholar]
  16. Futterer, J., A. Gisel, V. Iglesias, A. Kloti, B. Kost et al., 1995. Standard molecular techniques for the analysis of transgenic plants, pp. 215–217 in Gene Transfer to Plants, edited by I. Potrykus and G. Spangenberg. Springer, New York.
  17. Goodnight, C. J., 1988. Epistasis and the effect of founder events on the additive genetic variance. Evolution 42: 441–454. [DOI] [PubMed] [Google Scholar]
  18. Hallauer, A. R., and J. B. Miranda, 1988. Quantitative Genetics in Maize Breeding. Iowa State University Press, Ames, IA.
  19. Hill, W. G., and L. Bunger, 2004. Inferences on the genetics of quantitative traits from long-term selection in laboratory and domestic animals. Plant Breed. Rev. 24(II): 169–210. [Google Scholar]
  20. Hill, W. G., and A. Caballero, 1992. Artificial selection experiments. Annu. Rev. Ecol. Syst. 23: 287–310. [Google Scholar]
  21. Jaccard, P., 1908. Nouvelles recherches sur la distribution florale. Bull. Soc. Vaud Sci. Nat. 44: 223–270. [Google Scholar]
  22. Lopez-Fanjul, C., and A. Villaverde, 1989. Inbreeding increases genetic variance for viability in Drosophila melanogaster. Evolution 43: 1800–1804. [DOI] [PubMed] [Google Scholar]
  23. Lynch, M., and B. G. Milligan, 1994. Analysis of population genetic structure with RAPD markers. Mol. Ecol. 9: 91–99. [DOI] [PubMed] [Google Scholar]
  24. Maita, R., and J. G. Coors, 1996. Twenty cycles of biparental mass selection for prolificacy in the open-pollinated maize population Golden Glow. Crop Sci. 36: 1527–1532. [Google Scholar]
  25. Mayr, E., 1954. Changes in the genetic environment and evolution, pp. 157–180 in Evolution As a Process, edited by J. Huxley, A. C. Hardy, and E. B. Ford. George Allen & Unwin, London.
  26. Mayr, E., 1963. Animal Species and Evolution. Harvard University Press, Cambridge, MA.
  27. Naciri-Graven, Y., and J. Goudet, 2003. The additive genetic variance after bottlenecks is affected by the number of loci involved in epistatic interactions. Evolution 57: 706–716. [DOI] [PubMed] [Google Scholar]
  28. Nei, M., T. Maruyama and R. Chakraborty, 1975. The bottleneck effect and genetic variability in populations. Evolution 29: 1–10. [DOI] [PubMed] [Google Scholar]
  29. National Institutes of Health, 2002. Public domain image analysis software (http://rsb.info.nih.gov/nih-image/).
  30. Reverter, A., C. J. Kaiser and C. H. Mallinckrodt, 1998. A bootstrap approach to confidence regions for genetic parameters from method R estimates. J. Anim. Sci. 76: 2263–2271. [DOI] [PubMed] [Google Scholar]
  31. Robertson, A., 1952. The effect of inbreeding on the variation due to recessive genes. Genetics 37: 189–207. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Ruano, R. G., L. S. Silvela, C. Lopez-Fanjul and M. A. Toro, 1996. Changes in the additive variance of a fitness-related trait with inbreeding in Tribolium castaneum. J. Anim. Breed. Gen. 113: 93–97. [Google Scholar]
  33. Saccheri, I. J., R. A. Nichols and P. M. Brakefield, 2001. Effects of bottlenecks on quantitative genetic variation in the butterfly Bicyclus anynana. Genet. Res. 77: 167–181. [DOI] [PubMed] [Google Scholar]
  34. Vekemans, X., 2002. AFLP-SURV version 1.0. Laboratoire de Genetique et Ecologie Vegetale, Universite Libre de Bruxelles, Brussels.
  35. Vos, P., R. Hogers, M. Bleeker, M. Reijans, T. van de Lee et al., 1995. AFLP: a new technique for DNA fingerprinting. Nucleic Acids Res. 23: 4407–4414. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Whitt, S. R., L. M. Wilson, M. I. Tenaillon, B. S. Gaut and E. S. Buckler, 2002. Genetic diversity and selection in the maize starch pathway. Proc. Natl. Acad. Sci. USA 99: 12959–12962. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Williams, P. H., and C. B. Hill, 1986. Rapid-cycling populations of Brassica. Science 232: 1385–1389. [DOI] [PubMed] [Google Scholar]
  38. Willis, J. H., and H. A. Orr, 1993. Increased heritable variation following population bottlenecks: the role of dominance. Evolution 47: 949–957. [DOI] [PubMed] [Google Scholar]
  39. Wright, S., 1968. Evolution and Genetics of Populations, Vol I: Genetic and Biometric Foundations. University of Chicago Press, Chicago.
  40. Zhang, X.-S., J. Wang and W. G. Hill, 2004. Redistribution of gene frequency and changes of genetic variation following a bottleneck in population size. Genetics 167: 1475–1492. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Genetics are provided here courtesy of Oxford University Press

RESOURCES