The effect of inbreeding on phenotypic variance in the red flour beetle Tribolium castaneum.
The effects of inbreeding on phenotypic variance are well documented in the early evolutionary biology literature and reviewed by Lerner (1954), Wright (1977), and Falconer (1989) (see also Whitlock 1995). These early experiments provide mixed results. According to additive genetic theory, inbreeding removes genetic variance, thereby decreasing phenotypic variance (Falconer 1989). Inbreeding often increases an organism's sensitivity to environmental change, however, which may offset decreased genetic variance such that phenotypic variance actually increases (Falconer 1989). Even in a constant environment, phenotypic variance may increase due to "accidents of implantation or development" (Wright 1977). Moreover, evidence is accumulating to suggest that additive genetic variance, and consequently phenotypic variance, can actually increase following a period of inbreeding (e.g., Bryant and Meffert 1988; Bryant et al. 1986; Carson and Wisotzkey 1989; Goodnight 1987, 1988; Lopez-Fanjul and Villaverde 1989; Robertson 1952). The mixed results on the effects of inbreeding on phenotypic variance in the early literature presumably arise from variable changes in both external and internal, i.e., developmental, environmental sources of variation as a result of inbreeding, as well as variation in changes in genetic variance.
Most of the early studies compared inbred populations with their [F.sub.1] hybrids. Few studies have examined phenotypic variation among inbred populations relative to an outbred control, as opposed to an [F.sub.1] hybrid. Most recently, Whitlock and Fowler (1996) provide essential evidence for the distribution of within-population phenotypic variance within a single experiment and, with reference to the early experiments, suggest that "confusion of results is likely not due to differences between species or characters studied, but rather to the expected differences among lines in the response to inbreeding." In this paper, we provide further evidence on the distribution of phenotypic variation among populations across a range of levels of inbreeding (defined f = 0 and theoretically expected f = 0.25, 0.375, 0.5) achieved through a full-sibmating design.
MATERIALS AND METHODS
The methods used in this experiment are described for Experiment 2 in Pray and Goodnight (1995). Data were collected on 15 lineages derived from an ancestral Tribolium castaneum cSM + / + stock population. The cSM strain originated in 1973 in Michael J. Wade's lab at The University of Chicago, Chicago, Illinois, and is comprised of a mixture of four of Thomas Park's inbred strains. Beetles were maintained in standard flour medium (95% whole wheat flour, 5% dried brewer's yeast, 0.03% fumigillin by weight) at standard temperature (29 [degrees] C) and humidity (70% RH).
Each lineage was founded by a previously unmated parental pair randomly selected from the stock population (see Pray and Goodnight 1995 for a diagram of experimental design). The first data were collected on the offspring of the founder pairs (parent and offspring f = 0). Five full sib pairs from each lineage were randomly selected from these offspring to parent offspring for data analyses (offspring theoretically expected f = 0.25). From one of these sets of offspring, five full sib pairs were randomly selected to parent offspring for the next generation of data analyses (f = 0.375). Offspring from one of these pairs were selected to parent the final generation. Data were collected for a total of four generations (f = 0, 0.25, 0.375, 0.5).
We analyzed data for three of the four traits described in Pray and Goodnight (1995): egg-to-adult development time, proportion of offspring in a test population or "relative fitness," and adult dry weight. Data were collected on five females and five males from each of the five full sib pairs from each lineage each generation. Full sib pairs were allowed to mate and oviposit for 72 h. Five female and five male pupae were randomly selected from each full sib pair on the 28th day ([+ or -] 1 d) of development. Our estimate of [TABULAR DATA FOR TABLE 1 OMITTED] phenotypic variation is variation within a set of sisters or brothers per generation.
Assays for each trait are as described in Pray and Goodnight (1995): For egg-to-adult development time, each 28-day-old pupa was placed on 1 g standard media in a 15 x 45 mm vial and placed in racks positioned randomly in the incubator. Pupae were checked thereafter every 12 h and time of eclosion recorded. The same beetles were used in the proportion of offspring in a test population and adult dry weight assays.
For the proportion of offspring in a test population or "relative fitness" assay, the test population consisted of five females and five males. Nine of these were homozygous for a black body color marker (cSM b/b); the test individual was homozygous red (cSM +/+). All progeny of the test individual were heterozygous (+/b) brown and phenotypically distinguishable from the black progeny. The beetles were allowed to mate and oviposit for 11 d in a 25 x 95 mm shell vial containing 8 g standard media and were then removed. The offspring were collected and counted after 42 d from the start of the assay and the proportion of brown offspring calculated and used as an estimate of relative fitness. When the test individual in the relative fitness was removed from the test population on day 11, it was dried for at least 48 h in a 55 [degrees] C oven and weighed on a Cahn microbalance to the nearest 0.0001 mg to obtain an estimate for adult dry weight.
A panmictic (defined f = 0) control population was maintained for the duration of the experiment. For every generation, 25 pairs of outbred beetles were randomly selected from the control population to parent offspring on which the same assays as described above were performed.
Data for each sex were analyzed separately in the same manner as described for Experiment 2 in Pray and Goodnight (1995) with one important difference. The dependent variables for each trait were transformed using the r transformation described by O'Brien (1979, 1986) for testing variance equality. O'Brien's method provides a robust test of variance equality by simple transformation of the dependent variable. The transformed variable can be used in a regular analysis of variance. We transformed the control data before incorporating them into the analyses as covariates, as described by Muir (1986) (see also Pray and Goodnight 1995 and Wade et al. 1996). Muir's method corrects for uncontrolled environmental change and genotype by environment interaction between the control and inbred populations.
Data were analyzed on JMP (SAS 1989) in a two-way analysis of variance with lineage and the inbreeding coefficient f as the two main factors, as described in Pray and Goodnight (1995). We used the nested replicate (lineage, f) effect as an error term over which the lineage and lineage x f effects were tested. We tested for an overall increase in phenotypic variation by testing the linear contrast off. The significance of the lineage x f interaction was used to evaluate to what extent the effect of inbreeding on phenotypic variation is influenced by genetic lineage, and the linear component of the lineage x f interaction was used to test the extent to which the lineage effect is transmitted from generation to generation. Its significance indicates that the slopes of individual lineages vary, i.e., a steeper slope means a greater increase in phenotypic variance as a result of inbreeding. Another way of interpreting the linear component of the lineage x f interaction term is by estimating the "lineage heritability," or population heritability, of within-lineage phenotypic variance (Wade and McCauley 1980, 1984). We estimated lineage heritability as the square of the correlation coefficient, [r.sup.2], between generations (Wade and McCauley 1980, 1984). The correlation cefficient [r.sup.2] provides a measure of the proportion of total variance in phenotypic variance that is explained by among-lineage differences (Wade and McCauley 1980, 1984).
Finally, phenotypic correlations among the traits for each sex were analyzed in the same manner as described for phenotypic variances (a two-way ANOVA with phenotypic correlations in the outbred control incorporated into the analyses as covariates), except the correlations were not transformed. Unfortunately, the experimental design does not allow for estimation of individual-level genetic correlations, and population-level genetic correlations have been analyzed in another study (Pray, in press). Phenotypic correlations should, however, provide sufficient information to evaluate the dependency of phenotypic variances, even if information on genetic correlations is not available. For example, if heavier beetles are more fit, then phenotypic variance in weight is not independent from phenotypic variance in fitness, irrespective of the genetic correlation between weight and fitness.
There is a significant overall increase in phenotypic variation for three of the six traits - female fitness, male fitness, and female development time - as evident by a significant linear trend in f (Table 1; [ILLUSTRATION FOR FIGURE 1 OMITTED]). There is a significant linear trend in f for female adult weight, but it is negative (Table 1). Female relative fitness, female development time, and male relative fitness all show a significant lineage effect, suggesting that there is lineage-level variation in phenotypic variance for these traits. Furthermore, there is a significant linear lineage x f interaction component for male relative fitness (P = 0.0009), as well as male adult weight (P = 0.0006), suggesting that the lineage-level genetic component is transmitted across generations. For male weight, the significance of the linear interaction component suggests that even though on average there is no increase in phenotypic variation for this particular trait, for some lineages phenotypic variation does increase as evident in Figure 1.
One would expect male adult weight and male relative fitness to have significant lineage heritabilities because of their significant linear lineage x f interaction component. However, the lineage heritability estimates for within-lineage phenotypic variance in male adult weight are near zero. The lineage heritability estimates for within-lineage phenotypic variance in male relative fitness are consistently larger than the other traits (Table 2), but they are not statistically significant. The only statistically significant correlation coefficient is female development time, between the first (f = 0) and second (f = 0.25) generations.
The P-values for the linear lineage x f interaction components of the ANOVAs for phenotypic correlations are presented [TABULAR DATA FOR TABLE 2 OMITTED] in Table 3. All six correlations show a highly significant linear interaction, suggesting that there is significant variation in phenotypic correlations among lineages. Phenotypic correlations increase with increased inbreeding in some lineages, whereas they decrease in others.
There are two important conclusions that result from this study. First, phenotypic variance increases as a result of in-breeding for some traits. Second, for two traits (male relative fitness and male adult weight), there is variation in the response to inbreeding such that in some lineages phenotypic variance increases whereas in others it decreases, irrespective of the average change.
Our results corroborate reports by Whitlock and Fowler (1996) and Lopez-Fanjul and Villaverde (1989) that phenotypic variance can increase as a result of inbreeding, even in a constant environment. Whitlock and Fowler (1996) report a distribution of phenotypic variance following a bottleneck, for thorax length and sternopleural bristle number, in Drosophila melanogaster. They find that even though phenotypic variance decreases on average, some inbred populations can show a significant increase in phenotypic variance. Lopez-Fanjul and Villaverde (1989) also report an increase in phenotypic variance for egg-to-pupa viability in D. melanogaster.
For male adult weight and male relative fitness, a population's response to inbreeding with respect to phenotypic variation depends on genetic lineage as well as level of in-breeding. This is evident by their significant linear interaction components. Significant variation among lineages clearly shows that there is a lineage-level genetic component to trends in phenotypic variance. However, neither lineage heritability estimate is statistically significant (Table 2). There is probably too much within-lineage and within-replicate variation to detect statistically significant correlations at the lineage level. The proportion of variation explained by the lineage or lineage x f effect is very low (Table 1). Although the lineage x f interactions for female development time and female relative fitness are not significant, both of these traits show a significant lineage effect, suggesting that there is variation in phenotypic variance among lineages that does not depend on f.
TABLE 3. P-values for linear components of the lineage x f interaction in phenotypic correlation ANOVAs. A significant linear interaction component indicates that there is variation in phenotypic correlations among lineages. Phenotypic correlations increase with increased inbreeding for some lineages, but decrease for others. Correlation P-value Female fitness x weight [less than] 0.0001(***) fitness x development time 0.0007(***) weight x development time 0.008(**) Male fitness x weight [less than] 0.0001(***) fitness x development time 0.002(**) weight x development time 0.007(**)
The significant linear interaction components for the phenotypic correlation data suggest that phenotypic correlations do exist and they vary in magnitude and direction among lineages. In some lineages, phenotypic correlations increase significantly with inbreeding, whereas in others the correlations decrease. This is true for all six sets of correlations. If two traits are phenotypically correlated, then the analyses of their phenotypic variance are not independent. The observed increase in phenotypic variation for female fitness and female development time, for example, are not necessarily independent observations; however, their degree of dependence is not clear.
In summary, we have shown that phenotypic variance can increase as a result of inbreeding. Furthermore, phenotypic variance has a lineage-level genetic component for some traits. Increased phenotypic variation as a result of inbreeding can have implications for an inbred population's ability to respond to selection. This is especially relevant to conservation biology. One suggestion to eliminate inbreeding depression in captive populations of threatened and endangered species involves inbreeding followed by selection (Templeton and Read 1983; Backus et al. 1995). If an increase in phenotypic variation results from an increase in additive genetic variance, heritability can increase and thereby increase the potential response to selection. Wade et al. (1996) have shown that inbred beetles can in fact respond to selective pressures, even after five generations of full sib mating. Even if the increase in phenotypic variance results only from an increase in environmental variance, the potential response to selection may nonetheless increase by variance-induced peak shifts (Whitlock 1995). Finally, the lineage-level genetic component of phenotypic variance allows for the possibility of lineage-level selection for changes in phenotypic variance.
We especially thank M. Whitlock for suggesting that we examine the effect of inbreeding on phenotypic variance. We also thank M. Whitlock, L. Stevens, and especially M. Wade, for constructive comments and suggestions on the manuscript. This work was supported by NSF DEB-9307694 and NSF DEB-9321689.
BACKUS, V. L., E. H. BRYANT, C. R. HUGHES, AND L. M. MEFFERT. 1995. Effect of migration or inbreeding followed by selection on low founder number populations: Implications for captive breeding programs. Conserv. Biol. 9:1216-1224.
BRYANT, E. H., AND L. M. MEFFERT. 1988. The effect of an experimental bottleneck on morphological integration in the housefly. Evolution 42:698-707.
BRYANT, E. H., L. M. COMBS, AND S. A. MCCOMMAS. 1986a. The effect of an experimental bottleneck upon quantitative genetic variation in the housefly. Genetics 114:1191-1211.
CARSON, H. L., AND R. G. WISOTZKEY. 1989. Increase in genetic variance following a population bottleneck. Am. Nat. 134:668-673.
FALCONER, D. S. 1989. Introduction to quantitative genetics. 3d ed. Longman, London.
GOODNIGHT, C. J. 1987. On the effect of founder events on the epistatic genetic variance. Evolution 41:80-91.
-----. 1988. Epistasis and the effect of founder events on the additive genetic variance. Evolution 42:441-454.
LERNER, I. M. 1954. Genetic homeostasis. Oliver and Boyd, Edinburgh, Scotland.
LOPEZ-FANJUL, C., AND A. VILLAVERDE. 1989. Inbreeding increases genetic variance for viability in Drosophila melanogaster. Evolution 43:1800-1804.
MUIR, W. M. 1986. Estimation of response to selection and utilization of control populations for additional information and accuracy. Biometrics 42:381-391.
O'BRIEN, R. G. 1979. A general ANOVA method for robust tests of additive models for variances. J. Am. Stat. Assoc. 74:877-881.
-----. 1986. A simple test for variance effects in experimental designs. Psychol. Bull. 89:570-574.
PRAY, L. In press. The effect of inbreeding on population-level correlations. Evolution 51.
PRAY, L. A., AND C. J. GOODNIGHT. 1995. Genetic variation in inbreeding depression in the red flour beetle Tribolium castaneum. Evolution 49:176-188 .
ROBERTSON, A. 1952. The effect of inbreeding on variation due to recessive genes. Genetics 37:189-207.
SAS INSTITUTE. 1989. JMP user's guide. SAS Institute Inc., Cary, NC.
TEMPLETON, A. R., AND B. READ. 1983. The elimination of in-breeding depression in a captive herd of Speke's gazelle. Pp. 241-261 in C. M. Schonewald-Cox, S. M. Chambers, B. MacBryde, L. Thomas, eds., Genetics and conservation. Benjamin/Cummings, Menlo Park, CA.
WADE, M. J., AND D. E. MCCAULEY. 1980. Group selection: The phenotypic and genotypic differentiation of small populations. Evolution 34:799-812.
-----. 1984. Group selection: The interaction of local deme size and migration in the differentiation of small populations. Evolution 38:1047-1058.
WADE, M. J., S. M. SHUSTER, AND L. STEVENS. 1996. Inbreeding: Its effect on response to selection for pupal weight and the heritable variance in fitness in the flour beetle, Tribolium castaneum. Evolution 50:723-733.
WHITLOCK, M. C. 1995. Variance-induced peak shifts. Evolution 49:252-259.
WHITLOCK, M. C., AND K. FOWLER. 1996. The distribution among populations in phenotypic variance with inbreeding. Evolution 50:1919-1926.
WRIGHT, S. 1977. Evolution and the genetics of populations. Vol. 3. Univ. of Chicago Press, Chicago.
|Printer friendly Cite/link Email Feedback|
|Author:||Pray, Leslie A.; Goodnight, Charles J.|
|Date:||Feb 1, 1997|
|Previous Article:||Mating behavior in Drosophila melanogaster selected for altered longevity.|
|Next Article:||A cladistic analysis of pheromone evolution in Ips bark beetles (Coleoptera: scolytidae).|