Dr Juan L. Bouzat, Department of Biological Sciences, Bowling Green State University, Bowling Green, OH 43403-0212, USA. Tel.: (419) 372 9240; fax: (419) 372 2024; e-mail: email@example.com
An important issue in conservation biology and the study of evolution is the extent to which inbreeding depression can be reduced or reversed by natural selection. If the deleterious recessive alleles causing inbreeding depression can be ‘purged’ by natural selection, outbred populations that have a history of inbreeding are expected to be less susceptible to inbreeding depression. This expectation, however, has not been realized in previous laboratory experiments. In the present study, we used Drosophila melanogaster as a model system to test for an association between inbreeding history and inbreeding depression. We created six ‘purged’ populations from experimental lineages that had been maintained at a population size of 10 male–female pairs for 19 generations. We then measured the inbreeding depression that resulted from one generation of full-sib mating in the purged populations and in the original base population. The magnitude of inbreeding depression in the purged populations was approximately one-third of that observed in the original base population. In contrast to previous laboratory experiments, therefore, we found that inbreeding depression was reduced in populations that have a history of inbreeding. The large purging effects observed in this study may be attributable to the rate of historical inbreeding examined, which was slower than that considered in previous experiments.
It is widely agreed upon that most inbreeding depression in small populations is due to the expression of deleterious recessive alleles (Charlesworth & Charlesworth, 1999). This view of the genetic basis of inbreeding depression suggests that purging is a theoretical possibility, as in principle, the genes causing inbreeding depression may be eliminated from populations by selection during inbreeding. At present, however, there has been little empirical data to suggest that purging consistently plays a large role in small populations (Frankham et al., 2002). In two previous reviews of the empirical evidence, for example, it has been concluded that purging effects are generally small and insufficient to counter fitness declines during inbreeding (Ballou, 1997; Byers & Waller, 1999). In contrast, however, a third review by Crnokrak & Barrett (2002) found that most empirical studies provided evidence in favour of purging.
An important point that emerged from the review of Crnokrak & Barrett (2002) was that the design of experiments is an important consideration with respect to the detection of purging effects. A number of studies, for example, have shown nonsignificant fitness declines during inbreeding (e.g. Visscher et al., 2001; Fernandez et al., 2003), or fitness rebounds within inbred lineages over time (e.g. Bryant et al., 1990; Brakefield & Saccheri, 1994; Latter et al., 1995; Saccheri et al., 1996; Fowler & Whitlock, 1999). In all these cases, however, the extent to which purging occurred is not clear, since the maintenance or recovery of fitness during inbreeding can also be explained in terms of adaptation to the experimental or natural environment (Willis, 1999; Crnokrak & Barrett, 2002). The confounding effects of adaptation are avoided in experiments that test for an association between inbreeding history and inbreeding depression (e.g. Willis, 1999). In outbred populations that have a history of inbreeding, deleterious recessive alleles are more likely to have been at homozygous loci and thus exposed to elimination by natural selection. If purging effectively occurs, therefore, it is expected that outbred populations with a history of inbreeding will be less susceptible to inbreeding depression. This approach involves the comparison of inbreeding depression estimates rather than absolute fitness estimates between populations and therefore allows the effects of purging to be distinguished from those of adaptation.
The relationship between inbreeding history and inbreeding depression has been evaluated in several studies of natural plant populations (e.g. Latta & Ritland, 1994; Lacy & Ballou, 1998; Cheptou et al., 2000; Fishman, 2001; Haikola et al., 2001; Paland & Schmid, 2003), but has been examined in only a small number of laboratory experiments (Willis, 1999; Frankham et al., 2001; Radwan, 2003). The plant studies have generally tested for correlations between natural population sizes and the extent of inbreeding depression that occurs because of forced selfing. A potential drawback to this approach is that contemporary population sizes may not reliably reflect historical inbreeding. Several researchers have therefore advocated experimental approaches in which the inbreeding history of populations can be clearly established (Byers & Waller, 1999; Crnokrak & Barrett, 2002). Three previous laboratory studies have examined the relationship between inbreeding history and inbreeding depression using Mimmulus guttatus (Willis, 1999), Drosophila melanogaster (Frankham et al., 2001) and the bulb mite Rhizoglyphus robini (Radwan, 2003) as model systems. In all three of these cases, however, no significant relationship between inbreeding history and inbreeding depression has been found, suggesting that purging had not been effective during historical inbreeding.
In the present study, we used D. melanogaster to examine the relationship between inbreeding history and inbreeding depression. Similar to previous laboratory experiments (Willis, 1999; Frankham et al., 2001; Radwan, 2003), we compared estimates of inbreeding depression between populations that differ with respect to levels of historical inbreeding. In contrast to previous studies, however, we carried out historical inbreeding at a slow rate, i.e. at a larger population size over a greater number of generations. In previous laboratory experiments, historical inbreeding had occurred at a fast rate (either by selfing or full-sib mating). It is possible that such high rates of inbreeding limited purging effects (Ehiobu et al., 1989; Fu et al., 1998; Wang et al., 1999; Day et al., 2003), which may have decreased the likelihood of finding a significant association between inbreeding history and inbreeding depression. In our experiment, therefore, we created six ‘purged’ populations from a set of 30 lineages, which had been slowly inbred at a population size of 10 male–female pairs for 19 generations. We then measured the inbreeding depression that resulted from one generation of full-sib mating in the purged populations and in the original base population. Our primary objective was to address the hypothesis that natural selection removed deleterious recessive alleles during historical inbreeding and therefore reduced the susceptibility of purged populations to inbreeding depression.
The base population used in this experiment was founded from approximately 500 female D. melanogaster flies captured in August 2002 from an orchard located in Bowling Green, OH, USA. This population was subsequently maintained within 25 bottle cultures at a density of 25 male–female pairs per culture (N = 1250). Random mixing of individuals among the 25 bottle cultures was performed each generation throughout the experiment. The base population was used as a noninbred ‘control’ population, which was not subjected to purging of recessive lethals or mildly deleterious alleles. All laboratory populations were maintained at 25 °C on standard cornmeal-molasses food medium. Ether anaesthesia was used for the handling and transferring of flies.
A schematic illustrating the derivation of the purged populations is shown in Fig. 1. A set of 30 experimental lineages was derived from the base population after it had been maintained in the laboratory for 35 generations. Each of these lineages was initiated by a random base population sample of 10 male–female pairs. Lineages were then maintained at a population size of 10 pairs over the next 19 generations using randomly selected individuals as parents each generation. The inbreeding coefficient (F) of lineages at generation 19 was approximately equal to 0.50, assuming that the effective size of each lineage was 0.70 times the census population size (Wright & Kerr, 1954; Crow & Morton, 1955; Ehiobu et al., 1989). At generation 19, each of the 30 lineages was randomly assigned to one of six groups of five lineages each (Fig. 1). Within each group, 25 male–female pairs from each of the five lineages were randomly sampled, mixed and distributed uniformly among five culture bottles (five pairs per bottle). This yielded six ‘purged’ populations (designated A, B, C, D, E and F). Each purged population was noninbred (F ≈ 0), but possessed a history of inbreeding. Prior to the derivation of full-sib lines, the purged populations were each maintained at a population size of 125 male–female pairs within five culture bottles for the next three generations (with uniform mixing of randomly chosen individuals among the five culture bottles each generation).
Full-sib lines and fitness assays
A total of 180 full-sib lines were derived from the six purged populations (15 lines per purged population) and from the base population (90 lines) (see Fig. 1). Each full-sib line was established by a male and an unmated female that had been collected from different culture bottles of the base population or a given purged population. From the progeny of each male–female pair, two samples of five male–female pairs were randomly obtained and set up in two different bottle cultures (designated X and Y). Each sample of five pairs gave rise to inbred offspring (F = 0.25), which were collected from bottle cultures and used in the fitness assays.
Fitness was measured as the 72-h production of single females. This measure represents the number of offspring that emerged from single vial cultures after a female had been given 72-h to lay eggs. For each full-sib line, we evaluated the productivity of a cross between an individual emerging from a X bottle and an individual emerging from a Y bottle. This procedure ensured that the inbreeding coefficient was F = 0.25 in the parental pairs being assayed and the offspring that were counted. The productivity of 10 inbred females per full sib line was evaluated. In addition, for every 10 inbred females assayed, we simultaneously evaluated the production of 10 outbred females. As inbred and outbred females were measured simultaneously, the productivity of the outbred females served as an indicator of random environmental effects. This allowed environmental effects to be accounted for in estimates of inbreeding depression, without having to conduct fitness assays on all lineages simultaneously (Lynch, 1988). In the base population treatment, simultaneously measured outbred individuals were obtained from the base population, whereas in the purged population treatment, simultaneously measured outbred individuals were obtained from the particular purged population from which an inbred lineage was derived. This sampling scheme of outbred individuals served to account for possible differences in outbred fitness between the base population and purged populations when obtaining estimates of inbreeding depression.
The production assay was set up for each full-sib line by placing one female and two males into each of 20 separate culture vials. Of the 20 vials, 10 contained inbred females (each with two inbred males) from a single full-sib line and the other 10 held outbred females (each with two outbred males). These flies were initially held within the 20 vials for a period of 4 days to allow for mating. All flies were then simultaneously transferred to a new set of 20 vials for a 72-h period to allow the females to lay eggs. The parental flies were cleared from the second set of vials after 72 h and all offspring emerging from each vial were counted. This assay was carried out for all 180 full sib lines derived from the purged and base populations, such that the 72-h productivity of 3600 females (1800 inbred, 1800 outbred) was assayed throughout the whole experiment.
Our fitness assay yielded paired estimates of inbred and outbred fitness for each full-sib line that was measured. This allowed us to obtain inbreeding depression estimates that were independent of random environmental effects. The mean fitness of the inbred and outbred individuals corresponding to a single line are referred to as wI and wO, respectively. These values were used to calculate the inbreeding depression for each line (δ), which is equal to 1 − wI/wO (Lande & Schemske, 1985). The probability of inbreeding depression for each treatment was calculated by dividing the number of lines for which δ > 0 by the total number of lines derived from the base or purged populations. In addition, we report the average production estimates obtained across inbred and outbred lines within each treatment (i.e. I and O, respectively).
The replicate estimates of δ were not normally distributed for either experimental treatment. We therefore used nonparametric procedures to test whether median δ estimates were significantly greater than zero for each treatment and whether median δ estimates differed between treatments. The probability of inbreeding depression was compared between treatments using a chi-square test (Agresti, 1996).
Inbreeding depression in the base population
Among the 90 inbred lines derived from the base population, the individual mean 72-h production estimates (wI) varied considerably, ranging from 9.85 to 49.20. Averaged across all 90 inbred lines, the mean production estimate (I) was equal to 26.701 (SE = 1.178; see Table 1). Much of the variation in wI estimates was likely environmental, as similar levels of variation were observed in corresponding outbred production estimates (wO). Individual outbred production estimates (wO) ranged from 12.30 to 52.90, with an overall average (O) of 30.472 (SE = 1.38; Table 1). The correlation between inbred and paired outbred estimates was large and significant (rwI/wO = 0.860, P < 0.001), which suggested that outbred estimates were good indicators of random environmental effects. The overall median δ estimate of the base population treatment was equal to 0.145 (Table 1), which was significantly different from zero (Wilcoxon Signed Rank test; P = 0.001). The estimated value of δ was greater than zero in 70 of the 90 inbred lineages, indicating that the probability of inbreeding depression because of one generation of full sib mating in the base population was 77.8%.
Table 1. Mean 72-h production estimates of inbred and outbred females (I and O) measured in each of the n inbred lineages derived from the base and six purged populations.
Median estimates of inbreeding depression (δ) among the n lineages per treatment are listed.
Inbreeding depression in purged populations
The mean production estimates of inbred females in each of the six purged populations ranged from 25.537 to 40.601 (I) with an overall mean of 36.143 (SE = 1.097; Table 1). The corresponding outbred production estimates (O) ranged from 20.735 to 43.378 with an overall mean of 37.633 (SE = 1.482). Among the 89 inbred lineages derived from the six purged populations (one lineage was accidentally lost), the overall correlation between inbred and paired outbred production estimates was large and significant (rwI/wO = 0.808, P < 0.001). The median δ estimates of the six purged populations ranged from 0.145 to −0.364 (Table 1). Although this range is considerable, median δ estimates among the six purged populations were not significantly different (Kruskal–Wallis test; P = 0.076). When estimates of δ were pooled over all six purged populations, the overall median δ estimate was equal to 0.049 (Table 1), which was not significantly different from zero (Wilcoxon Signed Rank test; P = 0.6276). In 53 of the 89 individual estimates δ was greater than zero, indicating that the probability of inbreeding depression was 59.6% in inbred lineages derived from the purged populations.
Comparison of base and purged populations
The median level of inbreeding depression observed in lineages derived from the purged populations (δ = 0.049) was significantly less than that observed in lineages derived from the base population (δ = 0.145; Wilcoxon Signed Rank test; P < 0.001). This same result was obtained when a parametric procedure was used to compare the mean level of inbreeding depression in the purged populations ( = −0.033) to the mean estimate from the base population ( = 0.098; two-sample t-test; P < 0.001; following optimal box–cox transformation to approximate normality). The distribution of all 89 δ estimates associated with the purged populations relative to the distribution of 90 δ estimates associated with the base population is shown in Fig. 2a and box plots of the 95% confidence intervals associated with the treatment medians are shown in Fig. 2b. The probability of inbreeding depression did not significantly differ between the base and purged populations (χ2 = 1.287; P =0.257). However, the observed trend was consistent with the comparison of median δ values. The ‘relative risk’ of inbreeding depression, for example, was equal to 1.305 (77.8%/59.6%), which indicates that the probability of inbreeding depression was 30.5% greater in the base population than in the purged populations. There was also a significant treatment difference with respect to the production estimates associated with outbred individuals (Wilcoxon signed rank test; P = 0.003). Averaged across all six purged populations, the mean estimate of O was equal to 37.63 (SE = 1.48), whereas the estimate of O obtained from the base population was only 30.47 (SE = 1.38).
The results of this study have shown a significant reduction in the magnitude of inbreeding depression because of inbreeding history. The median level of inbreeding depression (δ) observed in the base population was equal to 0.145. In purged populations, the median level of inbreeding depression was approximately one-third of the base population value (0.049). These results are consistent with the hypothesis that natural selection removed deleterious recessive alleles from purged populations, which decreased levels of inbreeding depression in the purged populations relative to those of the base population. These findings corroborate evidence of purging effects from some previous studies (e.g. Bryant et al., 1990; Brakefield & Saccheri, 1994; Latter et al., 1995; Saccheri et al., 1996; Fowler & Whitlock, 1999; Visscher et al., 2001; Fernandez et al., 2003), but may represent the strongest evidence yet attained for the purging hypothesis because our approach prevented purging effects from being confounded by adaptation (Crnokrak & Barrett, 2002).
The experiments of Willis (1999), Frankham et al. (2001) and Radwan (2003) were similar in design to the present study, but did not demonstrate a significant relationship between inbreeding history and inbreeding depression. In each of these prior studies, extensive historical inbreeding had occurred, such that inbreeding coefficients >0.90 had been reached before purged populations were formed. In the present study, historical inbreeding was less extensive. The approximate inbreeding coefficients of lineages prior to the formation of purged populations was only 0.50. It is therefore surprising that historical inbreeding significantly reduced inbreeding depression in the present study but not in previous experiments. The nonsignificant purging effects in previous studies have not necessarily resulted from a lack of statistical power. Rather, the effect of inbreeding history appears to have been smaller. For example, the average reduction of inbreeding depression attributable to inbreeding history in the studies of Willis (1999), Frankham et al. (2001) and Radwan (2003) was only 14–19%, 7% and 21%, respectively. In the present study, inbreeding depression was reduced by 62% in the purged populations relative to the base population. The most likely explanation for the difference between our findings and those of previous experiments is the slow rate of historical inbreeding that we examined. Previous experimental studies have carried out historical inbreeding at a fast rate by selfing or full sib mating (Willis, 1999; Frankham et al., 2001; Radwan, 2003). Under such high rates of inbreeding, the influence of natural selection against deleterious recessive alleles may be extremely weak relative to the influence of genetic drift. Consequently, although historical inbreeding may have eliminated some recessive lethals in previous experiments (Hedrick, 1994), mildly detrimental recessive alleles may have been fixed within lineages and thus remained in purged populations. In the present study, therefore, although historical inbreeding was less extensive, a larger proportion of the genetic load may have been eliminated because of the slow rate at which historical inbreeding occurred. This explanation is consistent with the results of recent computer simulation studies (Fu et al., 1998; Wang et al., 1999), as well as the findings of several laboratory experiments (Ehiobu et al., 1989; Day et al., 2003; Reed et al., 2003), which have demonstrated reduced levels of inbreeding depression under slow inbreeding compared with fast inbreeding.
The fitness of outbred individuals from purged populations was significantly greater than that of outbred individuals from the original base population (see Table 1). This observation provides further evidence that substantial purging occurred during historical inbreeding (Crnokrak & Barrett, 2002; Roff, 2002). If purging occurs during inbreeding, crossing of inbred lines will yield outbred purged populations with a reduced equilibrium frequency of deleterious recessive alleles. This is expected to increase the outbred fitness of purged populations relative to the outbred fitness in the ancestral base population (Roff, 2002). As a consequence, the average level of fitness in populations with a history of inbreeding may exceed that of a population with no inbreeding history (Crnokrak & Barrett, 2002; Roff, 2002). Such increases in the average fitness of outbred individuals from purged populations have been observed in a number of previous experimental studies (e.g. Lynch, 1977; Bryant et al., 1990; Barrett & Charlesworth, 1991; Garcia et al., 1994; Latter et al., 1995; Willis, 1999). In the review of Crnokrak & Barrett (2002), the average increase in outbred fitness following a history of inbreeding was equal to 20% and 24% in mammal and plant populations, respectively. The estimated increase of 23% in the present study agrees well with these values. In some previous studies, adaptation to laboratory (or experimental) conditions may have been a source of bias when purging effects have been inferred from increases in outbred fitness in purged populations (Willis, 1999). This could not have been the case in our study because the purged populations and the base population had been maintained in the laboratory for the same number of generations (approximately 50) at the time of the fitness assays.
The purging effects observed in this study were because of selection against deleterious recessive alleles within lineages during historical inbreeding. Larger purging effects, however, may have been observed if selection had also occurred among lineages during historical inbreeding (Garcia et al., 1994; Goodnight & Stevens, 1996; Wang, 2000). All 30 lineages originally derived from the base population survived the 19 generations of slow inbreeding leading up to the formation of purged populations. After slow inbreeding, no among-lineage selection was performed, as we randomly grouped together and hybridized lineages in order to form the purged populations. Inbreeding depression and the effects of purging, however, are known to vary considerably among individual lineages (Pray & Goodnight, 1995; Lacy et al., 1996; Fowler & Whitlock, 1999). This variability, for example, is demonstrated by the wide distribution of δ estimates among lineages and the six purged populations (see Table 1 and Fig. 2a). As this variation provides substantial opportunity for among-lineage selection, it is likely that the effects of inbreeding history would have been greater if only those lineages with the highest fitness had been used to form a smaller number of purged populations. Periodic crosses among lineages during historical inbreeding might have enhanced purging effects still further. A recent computer simulation study (Wang, 2000), for example, revealed that historical inbreeding involving periodic line crosses, along with selection among and within lineages, is the most effective way of purging deleterious recessive alleles from populations. To our knowledge, no previous laboratory study has examined the relationship between inbreeding history and inbreeding depression under such a scheme of historical inbreeding.
The results of this study have implications regarding the genetic basis of inbreeding depression and the role of inbreeding depression within small populations of conservation concern. The finding of substantial purging because of historical inbreeding lends further support to the partial dominance hypothesis as an explanation for the genetic basis of inbreeding depression (Charlesworth & Charlesworth, 1999; Roff, 2002). The widespread acceptance of the partial dominance hypothesis as the primary explanation of inbreeding depression is somewhat conflicting with the view that purging is generally a weak force in small populations (e.g. Ballou, 1997; Byers & Waller, 1999; Frankham et al., 2002). When inbreeding occurs at a fast rate (e.g. Willis, 1999; Frankham et al., 2001; Radwan, 2003), it may be less likely that experiments will possess sufficient statistical power to detect treatment differences because of purging. The present study demonstrates, however, that significant purging effects may occur when inbreeding takes place at relatively slow rates. Although the purging effects we observed were strong, the inbreeding rate we examined was likely faster than that which occurs in natural populations (Ralls et al., 1986). It has been shown, for example, that rapid inbreeding caused by mating between first-order relatives is rare in natural populations (1–10% on average; see Keller & Waller, 2002), which may partly reflect inbreeding avoidance mechanisms that commonly influence mating patterns (Pusey & Wolf, 1996). In natural populations, therefore, inbreeding may occur at a slower rate than has been considered in this study, such that purging of deleterious recessive alleles may lessen the threat that inbreeding depression poses to population viability (e.g. Walter, 1990; Kalinowski et al., 1999; Visscher et al., 2001; Duarte et al., 2003).
This work was partially supported by National Science Foundation Doctoral Dissertation Improvement Grant Award No. DEB-0407891 and the Department of Biological Sciences and Bowling Green State University. We thank H. Michaels, D. D. Wiegmann and R. C. Woodruff for helpful advice and discussion. R. C. Woodruff is additionally acknowledged for providing working space and resources in his laboratory. We also thank L. Treeger and H. Strohschein for outstanding technical assistance.