SEARCH

SEARCH BY CITATION

Keywords:

  • bottleneck;
  • conservation;
  • Drosophila;
  • founder-flush;
  • inbreeding depression;
  • purging

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Drastic reductions in population size, or bottlenecks, are thought to significantly erode genetic variability and reduce fitness. However, it has been suggested that a population can be purged of the genetic load responsible for reduced fitness when subjected to bottlenecks. To investigate this phenomenon, we put a number of Drosophila melanogaster isofemale lines known to differ in inbreeding depression through four ‘founder-flush’ bottleneck cycles with flush sizes of 5 or 100 pairs and assayed for relative fitness (single-pair productivity) after each cycle. Following the founder-flush phase, the isofemale lines, with a large flush size and a history of inbreeding depression, recovered most of the fitness lost from early inbreeding, consistent with purging. The same isofemale lines, with a small flush size, did not regain fitness, consistent with the greater effect of genetic drift in these isofemale lines. On the other hand, the isofemale lines that did not show initial inbreeding depression declined in fitness after repeated bottlenecks, independent of the flush size. These results suggest that the nature of genetic variation in fitness may greatly influence the way in which populations respond to bottlenecks and that stochastic processes play an important role. Consequently, an attempt intentionally to purge a population of detrimental variation through inbreeding appears to be a risky strategy, particularly in the genetic management of endangered species.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

When a population passes through a generation with a small number of individuals, called a bottleneck, traditionally it has been thought that genetic drift results in a reduction in the overall heterozygosity, the number of different alleles, and the amount of additive genetic variation (James, 1971; Nei et al., 1975). For example, the low level of polymorphism observed for some genetic markers in cheetahs (O’Brien et al., 1983, 1985; Hedrick, 1996) and northern elephant seals (Bonnell & Selander, 1974; Hoelzel et al., 1993; Hedrick, 1995; Weber et al., 2000) has been attributed to bottlenecks in these species. The loss of genetic variation is also thought to have negative consequences for population fitness, both by fixing detrimental variants and reducing the genetic variation available for future adaptation.

On the other hand, bottlenecks have been suggested as a potential mechanism for the reorganization of genetic variation that could result in speciation (Carson & Templeton, 1984; however, see Barton & Charlesworth, 1984; Galiana et al., 1993). In addition, some experimental (Bryant et al., 1986; Carson & Wisotzkey, 1989; López-Fanjul & Villaverde, 1989; Fernandez et al., 1995) and theoretical (Robertson, 1952; Goodnight, 1987; Cockerham & Tachida, 1988; Willis & Orr, 1993) studies have suggested that bottlenecks may, in some instances, lead to increased additive genetic variance. However, generally in such cases the bottlenecks also result in reduction of mean fitness from the level found before the bottleneck and the conditions for an increase in additive genetic variance are quite restrictive (López-Fanjul et al., 1999).

Fitness is reduced in bottlenecked populations through random changes in frequencies for alleles influencing fitness. It has been suggested (Lande & Schemske, 1985) that populations may experience a smaller reduction in fitness over time. In other words, the set of detrimental alleles primarily responsible for lowered fitness from inbreeding or bottlenecks may be effectively purged by the combination of genetic drift and/or inbreeding and selection. For example, Templeton & Read (1983, 1984) suggested that the initial inbreeding depression for survival in a captive population of Speke’s gazelle (Gazella spekei) was reduced in animals whose parents were inbred (however, see Kalinowski et al., 2000), Bryant et al. (1990) reported that early inbreeding depression within serially bottlenecked isofemale lines of the housefly (Musca domestica) was largely ameliorated after later bottlenecks, and Saccheri et al. (1996) observed that a low hatching rate in a butterfly (Bicyclus anynana) caused by inbreeding was generally eliminated in later generations. However, Willis (1999) has recently suggested that it is difficult in laboratory studies to distinguish between purging of detrimental variants and adaptation unless the fitness of inbred individuals at the end of the experiment is evaluated.

The increase in subsequent fitness as the result of past inbreeding is not always apparent. For example, Ballou (1997) examined 25 captive species and found little effect of past inbreeding on fitness. A review of the plant literature by Byers & Waller (1999) found little evidence for purging among 52 studies. Moreover, the ability to purge a population of its detrimental variation appears highly dependent on the nature of that variation (Hedrick, 1994; Charlesworth & Charlesworth, 1999; Wang et al., 1999; Kirkpatrick & Jarne, 2000). Detrimental variation consisting primarily of recessive lethals is much more easily eliminated than that due to detrimentals with higher levels of dominance and lower effects.

In this paper, we report the results of experiments designed to investigate the effect of multiple bottlenecks on fitness in populations of Drosophila melanogaster. The general experimental protocol is similar to founder-flush studies of speciation (reviewed by Galiana et al., 1989; see also Galiana et al., 1993; López Bueno et al., 1993) in which bottlenecks (founder events) are alternated over time with generations exhibiting larger population numbers (flush periods). To gain insight into evolutionary dynamics and potential of the process, we examined populations that previously had either shown inbreeding depression or those that had not.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The isofemale lines used in this study were collected in Miami, Florida, by M. Tracey and were maintained in captivity for a minimal period of time before the initiation of the experiment. A population from south Florida was chosen because D. melanogaster populations in this area remain large throughout the year, with an associated high amount of detrimental genetic variation (Lewontin, 1974). A given isofemale line is established from the progeny of a single wild-caught female and includes at least four wild genomes (two from the female, two from a male) but may include more variation if the female was mated more than once. Once established, these isofemale lines were maintained at high numbers so little further inbreeding occurred. Examining different isofemale lines allows documentation of variation in a natural population for specific genetic characteristics (Hoffman & Parsons, 1988) without the problems associated with using long-term inbred lines. These isofemale lines and all experimental cultures were maintained in controlled conditions in environmental chambers with a 12-h photoperiod at 25 ± 1 °C under constant humidity (65 ± 5% RH) with the media consisting of measured amounts of yeasted Carolina Instant Medium.

Founder-flush protocol

Nineteen isofemale lines were initially subjected to three generations of single-pair full-sib mating (generations FS-1 to FS-3) in 8-dram vials, resulting in an inbreeding coefficient in the final set of progeny equal to f=0.50. Outbred controls were constructed by crossing virgin adults between the isofemale lines. More specifically, virgin adults of both sexes were randomly chosen and mated between the 19 isofemale lines for each generation of the three inbred generations, f=0.25, 0.375 and 0.5. For each of these generations, the control productivity (see below) was measured as the mean productivity of these crosses.

Following the final generation of inbreeding, 10 inbred isofemale lines were chosen for the founder-flush phase of the experiment: five isofemale lines that displayed considerable inbreeding depression relative to the control, and five isofemale lines showing comparatively little inbreeding depression relative to the control. This strategy allowed us to investigate the genetics of founder-flush dynamics in isofemale lines that differed substantially in their initial inbreeding depression. In particular, this protocol allowed us to examine both whether there was any improvement in fitness in those isofemale lines initially exhibiting inbreeding depression and whether isofemale lines initially having high fitness maintained their fitness.

For each of these 10 isofemale lines, two replicate half-pint culture bottles were set up for each of two flush treatments (40 total bottles). For both flush treatments, five pairs of adults were placed in each bottle on day 0, with adults removed three days later. On day 14, for the first flush treatment, called flush-5, five pairs were transferred to bottles containing new medium. For the second flush treatment, called flush-100, the first flush generation was the same as flush-5 while for the second generation, newly emerged adults up to a maximum of 100 pairs were similarly transferred to new bottles in each of the flush-100 lines. Each set of treatment bottles was maintained in this way for 7–10 generations, when a single-pair bottleneck was imposed on all lines (see Fig. 1). Following each bottleneck, the isofemale lines were again allowed to flush to their appropriate population sizes. All isofemale lines were taken through a total of four founder-flush cycles, designated FF-1 to FF-4, with the entire experiment encompassing 38 generations (76 weeks). The two flush treatments were intended to investigate the difference between a large flush size in which there is little potential for finite population size effects but opportunity for recombination and selection, and a low flush size in which genetic drift would be expected to dominate genetic change.

image

Figure 1.  Actual population sizes throughout the course of the experiment for the flush-5 and flush-100 isofemale lines. Fitness assays were conducted during the single-pair bottlenecks.

Download figure to PowerPoint

Fitness assessment

During the three initial generations of full-sib mating and each of the single-pair bottlenecks of the founder-flush cycles, fitness was measured as the total productivity of single pairs of adults in 8-dram vials. Productivity is a composite fitness measure, which includes female fecundity, male-mating success, and egg-to-adult survival. One of the two replicate bottles for each founder-flush line was chosen at random for the fitness measurement. Four replicate productivity vials were set up for each line during the continuous full-sib mating portion of the experiment, while five replicate vials were set up for each line in the two founder-flush treatments. Progeny were counted until day 17 for all vials, with the five pairs used to initiate the next set of flush bottles chosen at random from all available progeny.

The productivity control used in the fourth founder-flush cycle was generated by randomly choosing and mating flies from the 10 different flush-100 isofemale lines. We had planned to randomly choose and mate flies from stocks of the 10 selected isofemale lines for a control but stocks of several of these isofemales lines were lost over the 76-week experimental period. Because all 10 selected isofemale lines were represented in the flush-100 lines, we used crosses between them as a control because they provided the most complete genetic representation of the 10 selected isofemale lines available.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Following the final generation of continuous full-sib mating (f=0.5), inbreeding depression among the 19 inbred isofemale lines was shown by analysis of variance to be statistically significant (P=0.005). From these isofemale lines, the 10 most divergent were chosen to begin the founder-flush stage of the experiment: isofemale lines 3, 4, 12, 17 and 22 were designated as low-fitness isofemale lines, and isofemale lines 7, 15, 25, 30 and 31 as high-fitness isofemale lines.

Figure 2 shows the productivity data for the 10 selected isofemale lines during the continuous full-sib mating phase of the experiment. For the third generation of full-sib mating (f=0.5), the means (± standard error) for the control, low-fitness and high-fitness isofemale lines were 90.6 ± 9.9, 42.8 ± 12.5 and 83.3 ± 8.1, respectively. Analysis of variance of these data showed a significant effect of fitness group (P=0.017), with the low-fitness group showing significant inbreeding depression relative to the control (P < 0.05, Tukey’s a posteriori comparison) while the high-fitness isofemale lines did not (P > 0.50). Analysis of variance also showed that the mean productivity of the high-fitness group and the control during the continuous full-sib mating phase were not significantly different. Further, the control did not show significant variation over time, suggesting that there were no significant environmental trends or variation during this period.

image

Figure 2.  Productivity (mean ± SE) for the full-sib mating portion of the study. Data for the two fitness groups include only those nine isofemale lines used in the founder-flush protocol. Controls used during the experiment are also indicated.

Download figure to PowerPoint

The productivities of the flush-5 and flush-100 founder-flush isofemale lines and the controls at the beginning and end of the experiment are shown for the low-fitness (Fig. 3A) and high-fitness (Fig. 3B) isofemale lines. For the low-fitness isofemale lines, productivity declined from the initial value of 42.8 ± 12.5 in the flush-5 treatment to 29.5 ± 8.0 and increased in the flush-100 treatment to 48.1 ± 6.2 (line 4 was lost after the first founder-flush cycle). For the final founder-flush cycle, analysis of variance demonstrated that the flush-5, low-fitness isofemale lines exhibited significantly lower productivity compared to the control, while the flush-100 isofemale lines did not differ significantly from the control (Table 1).

image

Figure 3.  Productivity (mean ± SE) for the (A) low-fitness and (B) high-fitness groups over the four founder-flush cycles. Controls at the beginning and end of the experiment are also indicated.

Download figure to PowerPoint

Table 1.   Productivity (± SE) for the fourth and final founder-flush cycle where N is the number of test vials in each category (five replicates times the number of remaining lines for the experimental categories) and P is the significance level of each ANOVA comparing treatment means to the simultaneous outbred control. Thumbnail image of

Among the high-fitness isofemale lines (Fig. 3B), a fairly consistent decline in productivity was seen from the initial value of 83.8 ± 8.1 to 31.9 ± 3.7 and 29.3 ± 3.7 for the flush-5 and flush-100 treatments, respectively. The two treatments did not differ from each other in productivity for the final founder-flush cycle and both showed significantly lower productivity compared to the simultaneous control (Table 1).

Table 2 presents the productivity of each line, grouped by fitness and expressed as a proportion of their respective control values. For example, the FS-3 productivity for line 3 was 0.268 that of the control value and, using the Least Significant Difference (LSD) method for multiple planned comparisons (Sokal & Rohlf, 1995), it had significantly lower productivity compared to the simultaneous control. Isofemale lines 17 and 22 showed similarly low productivity in FS-3, but the differences were not quite statistically significant at the P=0.05 level. None of the five high-fitness isofemale lines shows significantly lower productivity relative to the simultaneous control in the FS-3 generation.

Table 2.   Productivity for each line expressed as a proportion of related control productivity for FS-3, the last generation of full-sib mating, and FF-4, after the last founder-flush cycle. Values indicated with *, ** and *** were significantly lower than the associated control values at the P < 0.05, 0.01 and 0.001 levels, respectively. – indicates loss of treatment line before the final productivity assay. Thumbnail image of

For the low-fitness isofemale lines in the FF-4 generation, line 3 showed significantly lower productivity in both the flush treatments, while line 17 had significantly lower productivity in the flush-100 treatment. However, average proportion over all low-fitness isofemale lines was unchanged for the flush-5 treatment while it was increased for all isofemale lines in the flush-100 treatment with a mean increase of 69%.

In marked contrast to these results, the high-fitness isofemale lines displayed much lower productivity following the FF-4 generation for both flush treatments. Isofemale lines 7, 25, and 30 showed significant reduction in productivity in both flush treatments and line 31 showed a similar reduction in the flush-100 treatment. As a result, the high-fitness isofemale lines as a group exhibit a significant reduction in fitness following the founder-flush protocol; average productivity in these isofemale lines was 92% of the control value after FS-3 and only 52% of the control at the end of the experiment.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Inbreeding, and its deleterious consequences for fitness, has been studied in considerable depth and is a topic of extensive current research (e.g. Charlesworth & Charlesworth, 1999; Bijlsma et al., 2000; Hedrick & Kalinowski, 2000). Nevertheless, the generality of the process by which the detrimental genetic variation of a population can be purged through inbreeding or bottlenecks remains a controversial area (Bijlsma et al., 1999; Byers & Waller, 1999; Willis, 1999).

Here we have observed that in the protocol design most likely to result in purging, such an outcome appeared in fact to occur. That is, in our examination of low-fitness isofemale lines (isofemale lines that had initially shown inbreeding depression) in the flush-100 regime (isofemale lines that were kept at 100 pairs between one-pair bottlenecks), the fitness relative to the simultaneous control increased from 0.473 to 0.801, an increase of 69%. For the low-fitness isofemale lines in the flush-5 regime (isofemale lines that were kept at 5 pairs between bottlenecks), fitness did not recover and was statistically indistinguishable from the initial level. We conclude that the higher flush numbers between bottlenecks in the flush-100 regime allowed selection, recombination, or perhaps mutation to regenerate the fitness initially lost from inbreeding. The absence of a fitness rebound for the flush-5 regime can be explained by the continued importance of genetic drift, even in the flush generations. However, because the fitness did not decrease in these isofemale lines further from the initial loss, even in the face of four one-pair bottlenecks, this suggests that the constraints caused by genetic drift appeared to be balanced by some effects of selection and recombination, or perhaps mutation, increasing fitness.

On the other hand, the high-fitness isofemale lines (isofemale lines that had not initially shown inbreeding depression) declined in fitness over time relative to the simultaneous control from 0.920 to 0.520, a decline of 57%. Potentially these isofemale lines could have increased in fitness if they had become more adapted to the laboratory environment (Willis, 1999). The decline was similar for both the flush-5 and flush-100 regimes, suggesting that the relative lack of genetic drift in the flush period for flush-100 regime did not overcome the effects of the four one-pair bottlenecks. We would have initially predicted that the high-fitness isofemale lines would have shown little further decline in fitness because the three generations of full-sib mating may have purged detrimental variants. On the contrary, these isofemale lines showed the greatest loss in fitness over the course of the experiment, independent of the flush size. A possible explanation is that the high-fitness isofemale lines were still segregating for detrimental genetic variants that were not expressed in the initial three generations of full-sib mating and that some of these variants became fixed in later bottlenecks.

Our results demonstrate the variance associated with the nature of fixation of the deleterious genetic variation in a population. Purely by chance, a population may become fixed or purged for detrimentals after just one bottleneck, or many bottlenecks may occur before the population experiences significant fixation or elimination of its deleterious genetic variation. One has no a priori knowledge of a particular population’s genetic response to either one or a series of bottlenecks. While there will always be unexplained variation in experimental studies of inbreeding depression that requires very large sample sizes to overcome (Lynch, 1988), the variation seen in our results may be explained by the stochastic nature of the change induced by population bottlenecks.

The results reported here might have important implications for application of evolutionary genetics principles in conservation biology. Templeton & Read (1983, 1984) concluded from their study in Speke’s gazelle that it may be possible to remove a large portion of the deleterious genetic variation in small captive populations of endangered species through deliberate selective inbreeding (see Kalinowski et al., 2000, for an alternative explanation). The rebounds in fitness reported by Bryant et al. (1990) and Saccheri et al. (1996) are also suggestive that inbreeding depression can be eliminated. However, from our study, and the reviews of Ballou (1997) and Byers & Waller (1999) it appears that a population’s response to bottlenecks is dependent not only on the population’s initial genetic constitution and the traits involved, but also on the chance genetic processes associated with bottlenecks. Application of the Speke’s gazelle breeding program to other animals may have quite different, and perhaps severe, consequences. Also, surveys have found that species differ widely in the severity of inbreeding depression (Ralls et al., 1988; for a review, see Hedrick & Kalinowski, 2000) and, in addition, theory predicts that a population of historically small size should have a lower equilibrium of lethal variants (Nei, 1968; Lande & Barrowclough, 1987; Bataillon & Kirkpatrick, 2000). As a result, one must expect the response of different species to population bottlenecks will vary considerably.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We thank Marty Tracey for collecting D. melanogaster samples for this research, Elizabeth King and Barb Webb for their technical assistance in the laboratory, and Ed Bryant for his comments on an earlier draft. This research was supported by NSF.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References