Tansley review. Environment-dependent inbreeding depression: its ecological and evolutionary significance. New Phytologist. Review

Transcription

1 Review Tansley review Environment-dependent inbreeding depression: its ecological and evolutionary significance Author for correspondence: Pierre-Olivier Cheptou Tel : +33 () pierre-olivier.cheptou@cefe.cnrs.fr Received: 26 July 21 Accepted: 27 September 21 Pierre-Olivier Cheptou 1 * and Kathleen Donohue 2 * 1 UMR 5175 CEFE Centre d Ecologie Fonctionnelle et Evolutive (CNRS), 1919 Route de Mende, F Montpellier, Cedex 5, France; 2 Department of Biology, Duke University, Box 9338, Durham, NC 2778, USA Contents Summary 395 I. Introduction 396 II. What is inbreeding depression? 396 III. Causes of environment-dependent inbreeding depression 397 IV. Ecological and evolutionary consequences of environment-dependent inbreeding depression 41 V. Feedbacks when inbreeding depression influences the environment 44 VI. Conclusions and future directions 45 Acknowledgements 46 References 46 Summary doi: /j x Key words: balancing selection, dispersal, environment dependence, evolutionary ecology, inbreeding depression, mating systems. Inbreeding depression is a major evolutionary and ecological force that influences population dynamics and the evolution of inbreeding-avoidance traits such as mating systems and dispersal. There is now compelling evidence that inbreeding depression is environment-dependent. Here, we discuss ecological and evolutionary consequences of environment-dependent inbreeding depression. The environmental dependence of inbreeding depression may be caused by environment-dependent phenotypic expression, environment-dependent dominance, and environmentdependent natural selection. The existence of environment-dependent inbreeding depression challenges classical models of inbreeding as caused by unconditionally deleterious alleles, and suggests that balancing selection may shape inbreeding depression in natural populations; loci associated with inbreeding depression in some environments may even contribute to adaptation to others. Environmentdependent inbreeding depression also has important, often neglected, ecological and evolutionary consequences: it can influence the demography of marginal or colonizing populations and alter adaptive optima of mating systems, dispersal, and their associated traits. Incorporating the environmental dependence of inbreeding depression into theoretical models and empirical studies is necessary for understanding the genetic and ecological basis of inbreeding depression and its consequences in natural populations. *These authors contributed equally to this work. 395

2 396 Review Tansley review New I. Introduction Darwin devoted an entire book to documenting inbreeding depression in plants (Darwin, 1876). His first motivation was to analyze the harmful effects of inbreeding, which he considered a major determinant of the adaptive value of outcrossing. Frequently overlooked is Darwin s observation that the intensity of these harmful effects of inbreeding depended on the environmental context. He wrote: The result was in several cases (but not so invariably as might have been expected) that the crossed plants did not exceed in height the self-fertilized in nearly so great a degree as when grown in pairs in the pots. Thus with the plants Digitalis, which competed together in pots, the crossed were to the self-fertilized in height as 1 to 7; whilst those which were grown separately were only as 1 to 85. Nearly the same result was observed with Brassica. With Nicotiana the crossed were to the self-fertilized in height, when grown extremely crowded together in pots, as 1 to 54; when grown much less crowded in pots as 1 to 66, and when grown in the open ground, so as to be subjected to but little competition, as 1 to 72 (Darwin, 1876). Darwin s observation of the environmental dependence of inbreeding depression has since been reported in diverse organisms (Miller, 1994; Keller & Waller, 22; Armbruster & Reed, 25; Fox & Reed, 21). More than 1 yr after Darwin, inbreeding depression is still of central interest in population biology. It is recognized as a major ecological and evolutionary factor that influences population demography and the evolution of traits, such as mating system and dispersal. Mating system and dispersal are fundamental determinants of population genetic structure, and they have long been recognized to be the primary factors that influence the rate and outcome of evolution (Wright, 1931). In part because of its influence on these traits, inbreeding depression itself has major evolutionary consequences. Extensive theoretical work has analyzed its genetic basis and evolutionary consequences (Charlesworth & Charlesworth, 1987; Charlesworth & Willis, 29), and this theory in turn has stimulated numerous empirical estimates of inbreeding depression (reviewed in Husband & Schemske, 1996; Byers & Waller, 1999; Keller & Waller, 22; Armbruster & Reed, 25; A. A. Winn et al., unpublished). As a consequence, inbreeding depression is an emblematic example in population biology of how theory and its confrontation with data have stimulated a vital field of research. Throughout much of this theoretical development, inbreeding depression was typically considered to be an intrinsic property of individuals, shaped by the history of inbreeding of organisms. Inbreeding depression is expected to vary with mating system as deleterious alleles are eliminated by selection (Lande & Schemske, 1985; Latta & Ritland, 1994) and it can vary among families as a result of stochastic genetic processes within populations (Dudash, 199; Wolfe, 1993; Dudash & Carr, 1998; Mutikainen & Delph, 1998; Culley et al., 1999; Moorad & Wade, 25). However, numerous empirical studies have revealed that the magnitude of inbreeding depression also varies greatly depending on the environmental context of the organisms (Fox & Reed, 21). Far from being simply a nuisance for the accurate estimation of the magnitude of inbreeding depression (Barrett & Harder, 1996), environment-dependent inbreeding depression (EDID) has important and often neglected ecological and evolutionary consequences (Lloyd, 198). It has become a focus of inquiry in itself, with empirical studies documenting its magnitude and nature of variation, and theoretical studies modeling its causes and ecological and evolutionary consequences. We review here the causes and consequences of EDID. We first present definitions of inbreeding depression and discuss potential causes of its environmental dependence. We then discuss ecological and evolutionary consequences of the environmental dependence of inbreeding depression, addressing its effects on population demography, matingsystem evolution, dispersal evolution, and the magnitude of inbreeding depression itself. II. What is inbreeding depression? Inbreeding depression is defined as the reduction in fitness of offspring resulting from matings between related individuals. This fitness reduction is considered to be caused by an increase in homozygosity across the genome resulting in the expression of unconditionally recessive deleterious alleles, a theory termed the partial dominance hypothesis, and or a reduction in heterozygotes for loci exhibiting unconditional heterozygote advantage, a theory termed the overdominance hypothesis (Charlesworth & Willis, 29). Most empirical evidence to date suggests that the majority of inbreeding depression is caused by deleterious recessive or partially recessive alleles that are manifest in the homozygous state in inbred individuals (Charlesworth & Charlesworth, 1987; Crow, 1993; Latter et al., 1995; Willis, 1999; Charlesworth & Willis, 29). A proportion of these recessive alleles appear to be lethal in the homozygous state, but a significant proportion appear to be nonlethal deleterious alleles, and partially recessive. Inbreeding depression is classically measured as the difference in fitness between inbred and outbred progeny, relative to the fitness of outbred progeny. As a particular case, an often used measure of inbreeding depression in hermaphroditic organisms is the relative decrease in fitness after one generation of selfing:

3 Tansley review Review 397 d ¼ðW outcrossed W selfed Þ=W outcrossed ¼ 1 ðw selfed =W outcrossed Þ; Eqn 1 where W selfed and W outcrossed are fitness estimates after selfing and outcrossing, respectively. More generally, it can be characterized as the rate at which fitness decreases with a given level of inbreeding, or the magnitude of an individual s inbreeding coefficient, f. A standard measure of inbreeding depression is the number of haploid lethal equivalents, B (Morton et al., 1956). B ¼ log e ðw inbred =W outbred Þ=f : Eqn 2 Inbreeding depression can be caused by alleles that are lethal when in the homozygous state or by alleles of smaller deleterious effect. The number of lethal equivalents measures the equivalent number of alleles that would cause lethality in the homozygous state. This measure is especially useful for populations with variation in the level of inbreeding, especially via biparental inbreeding. Measurements of mean inbreeding depression may be biased when the number of offspring is limited (Lynch & Walsh, 1997; Moorad & Wade, 25). Moreover, the relevance of relative, as opposed to absolute, measures of inbreeding depression has been challenged, as the mean and variance of relative measures of inbreeding depression are likely to be biased, even for an infinite number of offspring; as a consequence, absolute fitness values of inbred and outcrossed offspring have been argued to be more amenable to standard quantitative genetic models of evolutionary processes (Moorad & Wade, 25). However, most theory to date that models the effects of inbreeding depression on trait evolution (such as mating systems) has employed the relative measures defined by Eqns 1 and 2. In most treatments of inbreeding depression, loci that contribute to inbreeding depression are assumed to be invariably deleterious. Mutation selection balance models predict that deleterious alleles should be maintained at low frequency in populations, such that alleles associated with inbreeding depression are expected to be at low frequency. Several empirical studies, however, revealed that the frequencies of alleles responsible for inbreeding depression were too large to be explained by mutation selection balance alone (Charlesworth & Charlesworth, 1987; Kelly & Willis, 21; Charlesworth et al., 27; Charlesworth & Willis, 29). This result in itself questions the sufficiency of mutation selection balance to explain the maintenance of alleles associated with inbreeding depression; alleles causing inbreeding depression may not have consistently deleterious effects. While heterozygote advantage (the overdominance hypothesis) can maintain alleles at intermediate frequencies, it is not generally considered to be the major cause of inbreeding depression in many organisms (Charlesworth & Willis, 29). Antagonistic pleiotropy, moreover, can delay the erosion of genetic variation under mutation selection balance, if an allele has a favorable effect on one trait but a deleterious effect on another (Houle, 1992). Other forms of balancing selection, especially spatially variable environment-dependent natural selection, are effective at maintaining intermediate allele frequencies. The environmental dependence of inbreeding depression, coupled with the unpredictably high frequencies of alleles associated with inbreeding depression, suggests that loci that contribute to inbreeding depression may well be subject to balancing selection. Thus, the environmental dependence of inbreeding depression may significantly shape its genetic architecture. To a certain extent, understanding the maintenance of genetic variation in alleles contributing to inbreeding depression is closely linked to the more general question of the maintenance of genetic variation in fitness traits (Charlesworth & Willis, 29). III. Causes of environment-dependent inbreeding depression The strong effects of inbreeding depression on demographic and evolutionary outcomes have inspired numerous empirical studies of the magnitude of inbreeding depression. These studies revealed that the magnitude of inbreeding depression depends on the environmental conditions in which it was measured. By now, EDID has been reported in taxa as diverse as insects (e.g. Latter et al., 1995; Bijlsma et al., 2), fish (e.g. Gallardo & Neira, 25), plants (e.g. Schmitt & Ehrhardt, 199; Cheptou et al., 2), birds (e.g. Keller & Grant, 22), crustaceans (e.g. Haag et al., 22) and other taxa (reviewed in Miller, 1994; Byers & Waller, 1999; Keller & Waller, 22; Armbruster & Reed, 25). While several studies of previous decades emphasized that meaningful measurements of inbreeding depression must be made under natural conditions (see for instance Dudash, 199), it is imperative that 21st century studies of inbreeding depression pay particular attention to environmental conditions and consider them to be an important source of variation in inbreeding depression. Because of the desire to predict adverse effects of inbreeding in captive release programs or in populations already at risk, significant attention has recently been paid to whether effects of inbreeding are magnified in stressful conditions, when a stressful environment is defined as an environment that lowers absolute fitness. A recent survey of inbreeding depression studies in numerous taxa has indeed revealed that stressful environments often magnify inbreeding depression (Armbruster & Reed, 25; Fox & Reed, 21). However, other studies have reported lower inbreeding depression under stress (Henry et al., 23). Both outcomes are intuitive; the first results when inbred offspring are more

4 398 Review Tansley review New (a) Benign Stressful Favorable Stressful Fig. 1 Potential consequences of stress for the difference in fitness between inbred (rectangles) and outbred (diamonds) progeny. (a) Inbred progeny are more sensitive to stressful environments than outbred progeny, and the fitness difference between them increases in stressful environments. (b) Outbred progeny are more capable of capitalizing on favorable environments than inbred progeny, and the fitness difference between them is greater in the favorable environment. sensitive to environmental stress, and the second can result when outbred offspring are more capable of fully utilizing a favorable environment to increase their fitness (Fig. 1). Stressful conditions in experimental studies encompass a variety of biotic or abiotic environmental factors. Such diverse factors may be expected to have different effects on inbreeding depression, thereby precluding any strong pattern associated with stress generally. As a consequence, it is difficult to interpret inbreeding depression responses to stressful conditions in terms of general underlying mechanisms. To predict how the environment influences inbreeding depression, it is necessary to investigate the underlying mechanisms of EDID (Ågren & Schemske, 1993; Waller et al., 28). A number of mechanisms are possible, and they are likely to influence the ecological and evolutionary consequences of inbreeding depression differently. 1. Recognizing EDID The first step toward understanding the causes and consequences of EDID is to detect it. It is often assumed that a significant inbreeding environment interaction effect on fitness (i.e. the fact that the difference between inbred and outbred crosses changes with the environment) provides evidence for EDID. This is not necessarily true, depending on the definition of inbreeding depression that is of interest. EDID can be recognized either as an environmentally induced change in the absolute difference between the fitnesses of outbred and inbred progeny, or as an environmentally induced change in the relative fitness difference between outbred and inbred progeny, as defined in Eqn 1. Knowledge of the absolute fitness reduction caused by inbreeding is relevant for predicting demographic outcomes, such as decline in survival or fecundity, and their associated effects on population growth and extinction. In such contexts, it is relevant to analyse the absolute fitness of (b) individuals in populations with specified levels of inbreeding in specific environments. For inferring the evolutionary dynamics of different strategies that influence the probability of inbreeding (e.g. mating system or dispersal), evolutionary dynamics typically depend on the relative fitness of inbreeding vs outcrossing; thus, inbreeding depression defined as a ratio (as in Eqn 1) has been used most extensively (Lloyd, 1979; Lande & Schemske, 1985). More recently, however, it has been argued that absolute fitness measurements are more useful for predicting evolutionary outcomes with quantitative genetic models (Moorad & Wade, 25). If inbreeding depression is measured as the ratio of inbred to outcrossed fitness, this ratio may not differ across environments, even though the difference in fitness between inbred and outcrossed progeny varies and gives a significant inbreeding environment interaction effect on fitness (Fig. 2a). Also, a constant difference in absolute fitness between inbred and outbred progeny and an absence of inbreeding environment effects on fitness do not indicate the absence of environmental effects on the ratio (Fig. 2b). Moreover, a fitness estimate close to zero (e.g. survival) for both inbred and outbred progeny would lead to the conclusion of minimal inbreeding depression if it is measured as the absolute difference between inbred and outbred groups, whereas if it is measured as the ratio, it is mathematically undefined, and inbreeding depression may be either large or small (Fig. 2c,d). Note that the question of diverging inbreeding depression ratios has been addressed by Ågren & Schemske (1993), who proposed scaling the difference in fitness by the largest fitness value so that the ratio is bounded between þ1 and)1. The general implication is that a drastic reduction of absolute fitness (e.g. under high stress) is not likely to produce a consistent effect on the relative measure of inbreeding depression, but either higher or lower inbreeding depression can result compared with that observed in benign conditions. It is important to note, however, that natural log-transformed data would logically ensure that an inbreeding environment interaction can be interpreted as an effect of environment on inbreeding depression as measured by the ratio. The use of relative and absolute measurements of inbreeding depression is sometimes ambiguous in the empirical literature. It is important to clearly state the definition of inbreeding depression that is being used before reaching conclusions about its environment dependence. 2. Causes of EDID That inbreeding depression is affected by the environment has been empirically established, but the underlying mechanisms of EDID are far from clear. Theoretically, EDID can result from different mechanisms (Fig. 3). The difference in fitness between inbred and outbred progeny can change with the environment. First, this can occur when the envi-

5 Tansley review Review 399 W outbred W inbred (a) (b) (c) (d) Env. 1 Env. 2 Env. 1 Env Env. 1 Env. 2 Env. 1 Env. 2 Fig. 2 Upper panels: plots of the fitnesses of inbred (rectangles) and outbred (diamonds) progeny in two environments (Env ). Lower panels: changes in inbreeding depression measured as the absolute differences between outbred and inbred progeny (open circles), or the difference between outbred and inbred progeny divided by the fitness of outbred progeny (as in Eqn 1; closed circles). (a) The absolute difference in fitness differs across environments, but the ratio does not. A genotype (G ) environment (E) interaction for fitness does not indicate environment-dependent inbreeding depression (ID), as measured by the ratio in Eqn 1. (b) The absolute difference in fitness does not differ across environments, but the ratio does. No G E for fitness is apparent, even though ID, as measured by the ratio in Eqn 1, differs. (c, d) When one environment causes a reduction in fitness in both inbred and outcrossed progeny, the absolute difference in fitness is small in that environment, even though the ratio can be small or large. As the fitnesses of both inbred and outbred progeny approach zero, the ratio is mathematically undefined. In (c), the fitness of outbred progeny changed from 2 to 1, and the fitness of inbred progeny changed from 1 to.9. In (d), the fitness of outbred progeny changed from 2 to 1, and the fitness of inbred progeny changed from 1 to.1; W, fitness. (W outbred W inbred )/W outbred ronment influences the expression of phenotypes that are under selection (Fig. 3a). Even when selection on those phenotypes remains constant across environments, the fitnesses (and relative fitnesses) of inbred and outbred progeny can change if the environment alters the phenotypes that are expressed by those genotypes. Second, phenotypic expression may remain constant across environments, but selection on phenotypes may vary. This would also cause environment-dependent fitness differences between inbred and outbred progeny (Fig. 3b). This distinction between environment-dependent phenotypic expression and environment-dependent selection as causes of EDID is of biological interest. The first is caused by direct phenotypic responses to environmental conditions (plasticity), via changes in development, allocation or gene expression, and potentially implicates environmentally sensitive epigenetic processes in inbreeding depression variation (Biémont, 21). The second is caused by changing ecological conditions that determine the adaptive values of stable, nonplastic phenotypes. Thus, the genetic and ecological mechanisms differ fundamentally between those two scenarios. Third, the fitness difference between inbred and outbred progeny can change with the environment if dominance is environment-dependent (see for instance Bourguet et al., 1996). As heterozygotes result from outbreeding, the fitness of outbred progeny depends on whether the favorable or disadvantageous allele is dominant (Fig. 3c). Note that dominance can occur in fitness, when selection is stabilizing, even if alleles are completely additive with respect to phenotypic expression (Fig. 4). This can result in inbreeding depression even when effects of alleles on traits are purely additive (Lande & Schemske, 1985). The reason lies in the nonlinear relationship between phenotypic traits and fitness. Dominance relationships for fitness, moreover, can change with the environment if (Gaussian stabilizing) selection changes, specifically as a function of how close to the adaptive optimum the phenotypes are (Ronce et al., 29). Different genetic mechanisms, moreover, may underlie environmental differences in natural selection. For example, EDID may result from changes in the selection coefficients against deleterious recessive alleles, such that homozygotes for the recessive alleles are less deleterious in one environment than in another. Similarly, selection coefficients can range from deleterious to neutral, as would be the case if different loci contribute to inbreeding depression under different environments, as has been shown in empirical studies in Drosophila (Bijlsma et al., 1999). Porcher et al. (29) modeled such fluctuations of deleterious effects (selection coefficients) using the model of Kondrashov (1985). The Kondrashov approach models the evolution of the distribution of the number of homozygous and heterozygous deleterious mutations per individual in an infinite population. When selection coefficients vary temporally, these fluctuations shape the mean inbreeding depression as well as the variance of inbreeding depression expression at mutation selection equilibrium. Their results showed that corresponding fluctuations in inbreeding depression, as selection coefficients fluctuate, are predicted to be low. However, this predicted range of fluctuation in inbreeding depression was lower than that measured in published experimental studies

6 4 Review Tansley review New (a) A1 A1 Env. 1 Phenotype Env. 2 Env. 1 Env. 2 (b) Phenotype Phenotype Env. 1 Env. 2 Fig. 4 Alleles with additive effects on phenotypes can exhibit dominance for fitness under stabilizing selection. The dominance of fitness changes across environments as a function of the distance from the adaptive optimum, assuming a Gaussian fitness function. Well-adapted populations (close to the optimum; upper curve) are expected to exhibit more inbreeding depression than maladapted populations (far from the optimum; lower curve) as a consequence of the change in fitness curvature from accelerating to decelerating fitness increase. If stress is interpreted as maladaptation, inbreeding depression is predicted to be lower under stress than under benign conditions. Env., environment. (c) Phenotype A2 Env. 1 Env. 2 Fig. 3 Different genetic mechanisms for environment-dependent differences in fitness between inbred (rectangles) and outbred (diamonds) progeny. (a) The environment (Env.) induces changes in the phenotypes of inbred and outbred progeny, but selection on those phenotypes is constant across environments. (b) The environment does not influence phenotypic expression, but natural selection differs across environments. (c) Natural selection is constant across environments, and only dominance changes across environments. (Fig. 5), suggesting that a large amount of variation was not captured by this model, and that other mechanisms contribute to inbreeding depression variation. Empirical genetic analyses of inbreeding depression have concluded that loci responsible for inbreeding depression are often at higher frequencies than can be predicted by mutation selection balance alone (Charlesworth & Charlesworth, 1999; Kelly & Willis, 21; Charlesworth et al., 27; Charlesworth & Willis, 29). This result suggests that balancing selection operating at the scale of outcrossing distances may be maintaining alleles associated with inbreeding depression. When an allele is favorable in one environment but unfavorable in another, polymorphism can be maintained, even with migration (e.g. Epinat & Lenormand, 29). These alleles, moreover, can be said to contribute to local adaptation. Thus, this scenario of balancing selection is evolutionarily and ecologically fundamentally different from that of unconditionally deleterious alleles. Balancing selection on loci associated with inbreeding depression, however, has not yet been fully incorporated into theoretical models of inbreeding depression and its evolutionary consequences. When EDID is caused by fluctuations in selection, how well adapted a population is becomes important. In a recent model, Ronce et al. (29) showed that the magnitude of inbreeding depression can vary as a function of the distance between the population mean breeding value for a trait under stabilizing selection and the optimal phenotype. In their model, inbreeding depression is a function of the genetic variance for a trait under selection and the strength of stabilizing selection, as these factors influence both the mean relative fitnesses of inbred and outbred progeny and dominance (Fig. 4). The change in the dominance of fitness is caused by the change in curvature of the Gaussian fitness function. As genetic variance for the trait increases (with homozygotes of recessive alleles having more extreme values, for example), inbreeding depression also increases asymptotically. A major result of this model is that inbreeding depression is always lower when the population is less adapted to its environment compared with well-adapted populations. Under their assumptions, new environments are therefore expected to be associated with lower inbreeding depression, as are marginal populations that are impeded from adapting by gene flow of less adapted genotypes. The model of Ronce et al. (29) also demonstrated how

7 Tansley review Review 41 Variance of inbreeding depression Mean inbreeding depression U =.2 ; h =.2 ; s ² =.1 U =.2 ; h =.2 ; s ² =.3 U = 1 ; h =.2 ; s ² =.1 U = 1 ; h =.2 ; s ² =.3 U = 1 ; h =.3 ; s ² =.1 U = 1 ; h =.3 ; s ² =.3 Fig. 5 Predicted and empirically derived estimates of the mean and variance in inbreeding depression (redrawn from Porcher et al., 29). Inbreeding depression is described by Kondrashov s (1985) model, which models the evolution of the distribution of the number of homozygous and heterozygous deleterious mutations per individual in an infinite population for a given selfing rate. Mutation follows a Poisson process with rate U at an infinite number of loci. Fluctuating selection on deleterious mutations is modeled by sampling the selection coefficient of mutations, s, from a normal distribution truncated to the interval [,1], with mean l(s) and variance r 2 (s) before truncation. The selection coefficient is applied to all mutations, with a dominance coefficient, h. The fitness of an individual carrying x homozygous mutations and y heterozygous mutations is computed as (1 ) s) x (1 ) hs) y (multiplicative effects). Parameters of the simulations are listed in the figure. Empirical data (gray dots) are taken from Armbruster & Reed (25). The black envelope line represents the maximum variance in inbreeding depression, l(1 ) l), as a function of the mean inbreeding depression l (1 < l < ). Note that the variance in inbreeding depression covaries with the mean. inbreeding depression is affected by genetic variance. If inbreeding depression effectively purges extreme phenotypes over time, reducing genetic variance, then inbreeding depression is expected to decline and the interactions with environment are also expected to change, although these dynamics have not yet been modeled. The model of Ronce et al. (29) offers a clear interpretation of the relationship between stress and inbreeding depression: if we consider maladaptation (i.e. the distance from the optimal phenotype) to be equivalent to stress for an organism, we indeed expect that stress will actually decrease inbreeding depression (Fig. 4). Precisely how inbreeding depression changes with the environment depends on the genetic and ecological mechanisms of inbreeding depression. To date, we have little information on how prevalent those different mechanisms are. In particular, we do not know the relative contribution to EDID of environment-dependent gene expression determining particular phenotypes vs environment-dependent natural selection on stable phenotypes. When variable natural selection on phenotypes contributes to EDID, it could be manifest when the magnitude of purifying selection on specific loci is stronger in one environment than another, when different loci are under purifying selection in different environments, or when the loci are under balancing selection. It is frequently implicit in discussions of inbreeding depression that recessive homozygotes are unconditionally deleterious because of some intrinsic disruption of fundamental processes, similar to intrinsic Dobzhansky Muller incompatibilities that contribute to outbreeding depression. However, if the ecological context determines the adaptive value of loci associated with inbreeding depression, then balancing selection is likely, and indeed it is consistent with the empirical data on the genetic basis of inbreeding depression. Yet the role of balancing selection in the maintenance of alleles that contribute to inbreeding depression is still not known, nor are the ecological mechanisms of such balancing selection. This form of selection appears to have been greatly underappreciated in studies of inbreeding depression, given that the vast majority of models of inbreeding depression are based on the assumption of unconditionally deleterious mutational effects. The question of whether loci that contribute to inbreeding depression also contribute to local adaptation has seldom been posed, especially in empirical investigations of inbreeding depression. If loci that contribute to inbreeding depression also contribute to local adaptation, then the dynamics of adaptation will influence the evolution of inbreeding depression (Epinat & Lenormand, 29). This has important consequences for populations that are not at adaptive equilibrium, as a result of either dispersal or environmental change. To the extent that inbreeding depression influences population dynamics and trait evolution, ecological and evolutionary dynamics themselves will depend on how inbreeding depression changes with the environment and how inbreeding depression changes as populations become more or less adapted to their environments. IV. Ecological and evolutionary consequences of environment-dependent inbreeding depression 1. Effects of EDID on population demography Because of its effect on population growth and extinction probabilities, inbreeding depression is a subject of concern in conservation biology (Halley & Manasse, 1993; Frankham, 1995; Dudash & Fenster, 2; Hedrick & Kalinowski, 2; Luijten et al., 22). Contrary to arguments that inbreeding depression could be lower in the wild than in captivity (Frankham, 1995), inbreeding depression can be substantial in natural populations and has actually been shown to be higher in the wild (Crnokrak & Roff, 1999; Keller & Waller, 22). While density-dependent population regulation, moreover, has been argued to mask effects of inbreeding depression on population growth rates in natural populations, a recent simulation (Reed et al., 27) demonstrated that inbreeding depression can influence the probability of population extinction even with

8 42 Review Tansley review New density-dependent population regulation. Inbreeding depression has been shown to be caused not only by mating among close relatives but also by biparental inbreeding in small populations (Glemin et al., 23). Thus, inbreeding depression is significant in natural populations, and its effects on population viability are evident both theoretically and empirically. The environmental dependence of inbreeding depression is therefore likely to have important effects on population dynamics and stability. It is now clearly recognized that inbreeding depression estimated in laboratory or controlled conditions may not accurately reflect that expressed under natural conditions. As a consequence of EDID, estimates of inbreeding depression in the laboratory or controlled environments may give erroneous predictions about the viability of populations under natural conditions (Hedrick & Kalinowski, 2). With EDID, temporal variation in environmental conditions can cause fluctuations in inbreeding depression over time. To the extent that inbreeding depression affects population size, EDID can contribute to temporal variation in population size and consequently increase the risk of extinction. In a simulation, Liao & Reed (29) compared populations that exhibited temporal variation in inbreeding depression to those that exhibited a constant level of inbreeding depression that was equal to the mean of that expressed in the fluctuating populations. They found that temporal variation in inbreeding depression can lower effective population size and increase extinction risk. Interestingly, the effect of fluctuations on inbreeding depression was more pronounced in larger populations, suggesting that even populations that otherwise may not appear to be at risk of effects of inbreeding depression may exhibit some adverse consequences of it over time. In addition, a lineage environment interaction for inbreeding depression has been documented, indicating that EDID can change the relative fitness ranks of lineages (Pray et al., 1994). The theoretical prediction that inbreeding depression is lower in maladapted and marginal populations (Ronce et al., 29) is ostensibly in contrast to predictions that inbreeding depression will be more intense under stressful conditions (Armbruster & Reed, 25), if maladaptation is interpreted as stress. Distinguishing between these outcomes will be important for predicting the dynamics of populations undergoing environmental change or expanding their range and may be especially relevant for predicting the behavior of invasive species. If inbreeding depression is predictably lower in marginal populations, such populations may suffer less from inbreeding depression and may more easily evolve other traits, such as selfing, that could in turn affect their population structure and dynamics. Increased selfing in marginal populations has in fact often been reported in biogeographical studies (Stebbins, 1957; Randle et al., 29). These patterns are often interpreted as a consequence of mate or pollination limitation, but EDID in marginal habitats provides an additional explanation for the pattern (Ronce et al., 29). Empirical studies have seldom tested whether inbreeding depression varies with the degree of adaptation, although in a recent study, Pujol et al. (29) reported lower inbreeding depression in marginal populations than in central populations. One study of Impatiens capensis demonstrated that inbreeding depression increased with distance from the parental site, suggesting that, as local adaptation declines, inbreeding depression may be higher (Schmitt & Gamble, 199). Studies that explicitly test the degree of inbreeding depression as a function of the level of adaptation would be very useful for predicting the behavior of populations at risk of extinction and populations at risk of becoming invasive. 2. Effects of EDID on mating-system evolution As envisioned by Darwin (1876), inbreeding depression influences the evolution of outcrossing and selfing mechanisms, such as dioecy, heterostyly and cleistogamy (Lloyd, 1979; Bawa, 198). Theoretical models have demonstrated the fundamental role of inbreeding depression in determining the adaptive value of self-fertilization (Lloyd, 1979). The vast literature on how inbreeding depression affects mating-system evolution has provided a solid basis for investigating how EDID could alter those dynamics. In the context of the evolution of self-fertilization, for example, theory predicts a threshold value for inbreeding depression, such that when d <.5 selfing is advantageous and overcomes the 5% advantage of selfing genes over outcrossing genes (Fisher, 1941; Lloyd, 1979; Lande & Schemske, 1985). When d varies with spatial variation in the environment, outcrossing or selfing can therefore be favored in different environments. Cheptou & Mathias (21) demonstrated that temporal fluctuations of the environment lead to the maintenance of stable mixed selfing rates when EDID occurs. Fluctuating inbreeding depression greatly affected geometric fitness of full selfers, but partial selfing maximized fitness via bet-hedging. Although spatial heterogeneity could not maintain a mixed-mating strategy when dispersal occurred between environments, fluctuations in time could maintain a polymorphism of mixed strategies. This outcome of a stable mixed-mating system was not manifest in models that assumed a homogeneous and stable environment (Lloyd, 1979; Lande & Schemske, 1985). Thus, this model showed that EDID can facilitate the evolution of a stable mixed-mating strategy. A more recent model of the evolution of selfing with EDID included an explicit genetic mechanism of EDID and the possibility of purging of deleterious recessive alleles (Porcher et al., 29). In this treatment, EDID was generated when the magnitude of purifying selection varied across environments. A mixed-mating strategy was not

9 Tansley review Review 43 stable in this model, and the difference from the previous model appears to be to the result of two processes. First, under the Porcher et al. (29) genetic model of EDID, the mean inbreeding depression covaried with the variance in inbreeding depression (Fig. 4), such that low inbreeding depression tended to occur with low variance in inbreeding depression; with low variance in inbreeding depression, the cost of fluctuating inbreeding depression is lower, and both low inbreeding depression and low variance in inbreeding depression would operate to favor selfing. Secondly, purging appeared to cancel the effect of variance in inbreeding depression: even though the magnitude of inbreeding depression decreased with increased selfing, which would tend to favor outcrossing, the variance in inbreeding depression also decreased, which would favor selfing. To date, this is the only model of the effects of EDID on mating-system evolution that includes a genetic mechanism of inbreeding depression and purging. While it predicted no stable mixed-mating strategy, it should be noted that other genetic mechanisms of inbreeding depression (such as balancing selection) may produce different results. As a result, more theoretical and empirical work is needed, both on the genetic basis of inbreeding depression and on the conditions that influence the effectiveness of purging in natural populations. 3. Effects of EDID on dispersal evolution Like mating system, dispersal can influence the probability of inbreeding, and as a consequence its evolution is expected to depend on the magnitude of inbreeding depression. A reduction of inbreeding occurs with increased dispersal when populations are genetically structured (Levin, 1977, 1984; Fenster, 1991), so dispersal can be favored as an inbreeding-avoidance mechanism. Inbreeding depression alone favors dispersal (Bengtsson, 1978; Morgan, 22), but the costs of dispersal and outbreeding depression may overcome that advantage (Wiener & Feldman, 1993; Gandon, 1999; Roze & Rousset, 25). Inbreeding depression influences the evolutionary dynamics of dispersal not only by decreasing the fitness of nondispersed individuals that consequently inbreed, but also by altering population structure eliminating individuals with high inbreeding coefficients and thereby reducing the degree of relatedness among individuals within demes. This reduction of intrademic relatedness caused by inbreeding depression reduces competition among relatives, which in turn reduces the advantage of dispersal and favors nondispersal; this reduction of kin competition, however, overcomes the advantage of inbreeding avoidance via dispersal only when the cost of dispersal is high (Roze & Rousset, 25). In addition, inbreeding avoidance mechanisms such as mating system and dispersal potentially coevolve, and outcomes depend on the magnitude of inbreeding depression combined with all other factors that determine the adaptive value and costs of each trait (Ravigne et al., 26). Explicit genetic models of inbreeding depression have been incorporated in models of dispersal, and these include inbreeding depression caused by deleterious partial recessives (Roze & Rousset, 25) as well as heterozygote advantage (Wiener & Feldman, 1993). To summarize a vast theoretical literature on the evolution of dispersal, increased dispersal is favored by inbreeding depression alone (see the previous paragraph), kin competition (Hamilton & May, 1977) and temporal variation in environmental quality (e.g. Hastings, 1983). Dispersal is disfavored by a cost of dispersal, spatial heterogeneity in environmental quality, and local adaptation (Balkau & Feldman, 1973). The optimal dispersal strategy depends on the balance of all these factors. We expect that two processes may be especially likely to be affected by the environmental dependence of inbreeding depression. The first is the role of temporal and spatial variation in environmental quality in favoring dispersal. Dispersal reduces the risk of extinction of a lineage in uncertain environments by reducing the temporal variation in fitness when offspring are dispersed across a variable habitat. Reduction of temporal variation in fitness raises the rate of increase of that lineage and increases the effective population size of populations (Levins, 1968; Venable & Lawlor, 198; Hastings, 1983; Levin et al., 1984). EDID can also influence temporal variation in population size (Cheptou & Dieckmann, 22). Therefore, if inbred individuals suffer even more temporal variation in fitness, then dispersal as bet-hedging may be more strongly favored in inbred individuals. The second process is the role of spatial environmental variation and local adaptation (outbreeding depression) in favoring reduced dispersal. It is within this context that EDID may be especially relevant and its genetic basis highly pertinent. One model of the evolution of dispersal as a consequence of inbreeding depression has incorporated spatial variation in selection against homozygotes, using an overdominance mechanism for inbreeding depression (Wiener & Feldman, 1993). In a two-environment context, selection against each homozygote varied across environments, such that loci accounting for inbreeding depression were directly associated with local adaptation in a heterogeneous environment. In this model, dispersal was not easily favored, as outbreeding depression or the disruption of local adaptation overwhelmed the benefits of inbreeding avoidance, and a mixed-dispersal strategy sometimes resulted. While overdominance may not reflect the genetic basis of inbreeding depression most accurately for many organisms, this model is innovative in that it explicitly assesses how spatially variable selection on loci associated with inbreeding depression influences the evolution of inbreeding-avoidance traits. Incorporating local adaptation into models of dispersal as an inbreeding-avoidance mechanism, with explicit genetic mechanisms for inbreeding depression, seems a highly

10 44 Review Tansley review New promising avenue for future theoretical work (see Fenster & Galloway, 2 for an empirical study). Considering the evolution of dispersal as influenced simultaneously by inbreeding depression and local adaptation raises important questions concerning the relationship between loci associated with inbreeding depression and those associated with local adaptation. Given that local adaptation favors nondispersal and inbreeding depression favors dispersal, whether genes associated with inbreeding depression are the same genes associated with local adaptation (as opposed to being unconditionally deleterious, or ranging from deleterious to neutral) would probably have an influence on evolutionary outcomes. Moreover, the genetic basis of adaptation itself is relevant for predicting outcomes; specifically whether the same (tradeoffs and balancing selection on loci) or different loci contribute to local adaptation. Whether EDID is caused by balancing selection that variously favors and disfavors alleles, or whether it is caused by the contributions of different loci to inbreeding depression in different environments is also relevant. Thus, the effect of EDID on the evolution of dispersal is likely to offer an especially rich theoretical arena that engages fundamental issues of the genetic basis of adaptation, the genetic basis of inbreeding depression, and the effects of inbreeding depression on trait evolution. 4. Effects of EDID on purging The magnitude of inbreeding depression itself is expected to evolve as a result of the coevolution between the genetic architecture of inbreeding depression and inbreeding history; deleterious alleles associated with inbreeding depression are expressed as homozygotes more often when inbreeding is prevalent, and as such they can be more quickly eliminated by natural selection. This phenomenon, known as purging, has been intensively studied in theoretical models (Lande & Schemske, 1985; Charlesworth et al., 199). Empirical evidence for purging has been mixed, however (Husband & Schemske, 1996; Byers & Waller, 1999; Johnston et al., 29), with no clear relationship between the frequency of past inbreeding and the magnitude of present inbreeding depression. Some genera have indeed revealed inbreeding depression values consistent with a purging process (Carr et al., 1997; Dudash & Carr, 1998) while others have not (Johnston & Schoen, 1996). At the population level, individual history may have important consequences on among individual inbreeding depression variation (Shultz & Willis, 1995). The environmental dependence of inbreeding depression has important consequences for the evolution of inbreeding depression and the purging process, and may contribute to the lack of a strong association between inbreeding history and inbreeding depression. If the detrimental effects of deleterious alleles change with the environment, their frequency will change as a consequence, and environment-dependent purging should occur. Without consideration of the environmental dependence of inbreeding depression, predictions about how inbreeding will affect inbreeding depression will probably be inaccurate. That environment-dependent purging can occur has been supported by recent experimental studies (Bijlsma et al., 1999). The manner in which EDID affects the purging process is likely to depend on precisely how selection on loci associated with inbreeding varies across environments. For example, if selection on such loci is balancing, as opposed to unconditionally directional, then purging will be impeded. The degree to which it will be impeded will depend on the strength and direction of selection on loci associated with inbreeding depression in each environment, and on the frequency of exposure to each environment. EDID is therefore expected to influence the dynamics of purging, and it will also influence the persistence of new mutant alleles that contribute to inbreeding depression. The properties of genes that cause inbreeding depression (genomic mutation rates, dominance, and selection coefficients) have typically been investigated with little concern for the heterogeneity of selection. Similarly, estimates of mutational effects from mutation-accumulation lines have usually minimized environmental variation and so do not quantify environment-dependent mutational effects (Schoen, 25). However, some recent exemplary studies have estimated genetic parameters of new mutations in the field and demonstrated that they differ substantially from those estimated under laboratory conditions (Rutter et al., 21). Considering the environmental dependence of mutational effects and their contributions to inbreeding depression will provide a better understanding for the role of mutations in ecology. Moreover, as purging influences the evolutionary dynamics of inbreeding-avoidance traits such as mating system and dispersal, incorporating effects of EDID on purging is likely to alter predictions of evolutionary trajectories and outcomes. V. Feedbacks when inbreeding depression influences the environment Inbreeding depression has the potential to alter the environment that organisms experience. In particular, the competitive environment can be altered by inbreeding depression, either by reducing the number of competitors, or by altering the frequency of inbred vs outcrossed competitors. When this environment in turn alters inbreeding depression, feedbacks can occur. These environmental factors have, in fact, been shown to affect the magnitude of inbreeding depression. Competitive interactions (density dependence) and the degree of inbreeding of competitors (frequency dependence) both affect the magnitude of inbreeding depression. For instance,