Title: Maternal effects and the population dynamics of insects on plants
Abstract: If, like me, your mother continues to interfere in your life, far beyond her genetic contribution to your phenotype, then you may be sympathetic to the concept of maternal effects and their influence upon organismal ecology. Non-Mendelian maternal effects describe the transfer of information from the maternal environment to the phenotype of offspring (Mousseau & Fox, 1998). In essence, many aspects of the parental environment (climate, food quality, natural enemies) can influence the allocation of resources by parents to offspring and the quality of parental care. Before you scream ‘Lamarck’ and run for your copy of The Selfish Gene (Dawkins 1976), non-Mendelian maternal effects do not result in heritable (genetic) change in offspring. Rather, they describe the contribution of the parental environment to phenotypic variation in progeny. What does this have to do with the population ecology of agricultural and forest pests? My interest in maternal effects was reawakened recently at the combined meeting of the Royal Entomological Society and the International Union of Forestry Research Organizations in Aberdeen, Scotland (September 2001). At the meeting, I heard two presentations in which the expression of maternal effects could be inferred to explain patterns of population change in pest insects. In the first presentation, Werner Baltensweiler described the analysis of 50 years of data on the larch budmoth, Zieraphera diniana (Baltensweiler & Rubli, 1999). The larch budmoth has been considered to provide a classic case of population cycles driven by insect-mediated changes in plant quality (Baltensweiler, 1984; Baltensweiler & Fischlin, 1988). Declines in needle size and nitrogen concentration during budmoth outbreaks have been thought to provide the time-lagged negative feedback necessary to drive population cycles. However, at the conference, Baltensweiler reported that it may not be that simple. Towards the end of his 50-year dataset, reductions in needle size were not apparent during outbreaks, yet cycles continued as before. Having previously discounted the influence of natural enemies on population cycles, Baltensweiler suggested that time-lagged maternal effects may be responsible for cyclic dynamics. In a second presentation, Vince Nealis reported that the overwintering survival of spruce budworm (Choristoneura fumiferana) larvae was a major factor in population declines of budworm at the end of outbreaks. Budworm larvae hatch in the autumn and generally do not feed before entering winter diapause (Royama, 1992). Consequently, their resources for surviving the winter are provided entirely by the maternal provisioning of eggs. Nealis noted that declines in winter survival were associated with changes in the species composition of forest stands from which one might conclude (at least, I did) that changes in maternal diet may be responsible for changes in egg provisioning and subsequent declines in larval survival. The tremendous effort that has gone in to the collection of long-term data on spruce budworm and larch budmoth has yielded tantalizing glimpses of the possible action of maternal effects in pest population dynamics. Because few of us have collected population datasets even approaching the length of these classics, we have to hope that there exist faster ways to assess the prevalence of maternal effects in our own systems of study. Here, I explore what we know to date about the action of maternal effects and their potential to influence pest population dynamics. Maternal effects are ubiquitous in natural and managed ecological systems (Bernardo, 1996; Mousseau & Fox, 1998a). Organisms as diverse as baboons (Alberts, 1994), barnacles (Holm et al., 2000) and salamanders (Collazo, 1993) all exhibit maternal effects on offspring performance. The term ‘maternal effect’ is generally used as a shorthand for non-Mendelian parental effects, which can include contributions from the father as well as from the mother. For example, environmentally determined variation in the quality of male ejaculate as a resource for females would be considered as a paternal effect (Savalli & Fox, 1998) if it influenced the phenotype of offspring. For simplicity, I will stick with current convention and use the phrase ‘maternal effect’ to describe all non-Mendelian parental effects of both maternal and paternal origin. The forms and consequences of maternal effects are remarkably varied. A few of my favourites from the recent literature include maternal effects upon the size of cultured edible snails (Dupont-Nivet et al., 1997) and the glancing rate of infant baboons (Alberts, 1994). I was also amused to discover that both hormones and vaginocervical stimulation are required for the induction of maternal behaviour in sheep (Kendrick & Keverne, 1991). Of course, not every trait expressed by offspring is influenced by maternal effects. I am happy to report that male licking behaviour does not appear to have a maternal component, at least in Drosophila (Welbergen & Van Dijken, 1992). However, many of the traits that we do know to be important for population dynamics (Rossiter, 1991a, 1992, 1994; Ginzburg & Taneyhill, 1994; Benton et al., 2001) and rates of evolutionary change (Bernardo, 1996; Rossiter, 1996, 1997; Mousseau & Fox, 1998a; Wolf et al., 1999; Brodie & Agrawal, 2001) have strong maternal effects components. These include offspring survival, dispersal, growth rate, diapause and fecundity. There is also growing evidence that some maternal effects are not simply the accidental transmission of environmental information from one generation to the next. Rather, in many cases, the form and function of maternal effects appear to have been shaped by natural selection (Mousseau & Fox, 1998b). In many organisms, a female's environment provides a reliable indicator of the environmental conditions that her progeny will encounter. In such cases, maternal effects may evolve as mechanisms for transgenerational phenotypic plasticity, whereby, in response to a predictive environmental cue, a mother can change the type of eggs that she makes or can programme a developmental switch in her offspring, which produces offspring prepared for the environmental conditions predicted by the cue (Fox et al., 1999a). The first to recognize the potential for maternal effects to influence population dynamics were Wellington (1957) and Leslie (1959). Both pointed out that the phenotypic traits expressed by individuals within populations could vary among generations, leading to variation in susceptibility to environmental challenges. For example, Wellington (1957) tracked variation in the behaviour and performance of western tent caterpillars, Malacosoma pluviale, through different stages of population growth and concluded that dynamics were influenced by the changing quality of individuals. He developed this theme in subsequent work (Wellington, 1960, 1964) and came to the conclusion that population biologists ‘sometimes act as though they … have forgotten the animals from which their disciplines sprang’ (Wellington, 1977). In other words, populations are not lumps of phenotypically identical protoplasm that remain invariant within and among years. Maternal effects provide a mechanism by which the environment in a given year is expressed in the phenotypic variation of offspring in subsequent years (Rossiter, 1994) and therefore provide a route of delayed negative feedback (Berryman, 1999). It is well established that delayed density dependence can destabilize population dynamics and promote cyclic dynamics (Schaffer & Kot, 1986; Turchin, 1990; Royama, 1992). Exactly how is environmental information transferred between parental and offspring generations? One mechanism by which females influence the phenotype of their progeny is through heterogeneity in egg quality, in which females vary egg size and egg provisioning (Rossiter, 1991a,b; McIntyre & Gooding, 2000). Mothers who provide significant resources to their offspring by provisioning their eggs are essentially transmitting components of their own environment to that of their offspring in an oocytic ‘packed lunch’, and differential allocation by mothers of resources to eggs has become a classic example of a maternal effect (Rossiter, 1991a,b; Gliwicz & Guisande, 1992; Rossiter et al., 1993; Rolff, 1999). In some cases, the transmission of maternal environment to the offspring may simply reflect the non-adaptive legacy of previous conditions. Overall reductions in egg provisioning or larval performance when mothers are reared in poor environments provide examples of this type of maternal effect in insects (Gould, 1988; Jann & Ward, 1999; McIntyre & Gooding, 2000). The tragedy of ‘crack babies’ provides an example from humans. Nevertheless, maternal effects are not always the passive transmission of good or bad parental experiences to offspring. Egg provisioning in many species is a plastic trait in which environmental conditions influence the expression of the trade-off between egg size and egg number. In such cases, adaptive changes in resource allocation to offspring under varying environmental conditions can be considered as a form of flexibility in ‘family planning’. For example, Daphnia mothers grown under conditions of low food availability produce a few large offspring that are able to withstand some degree of starvation. In contrast, mothers grown under conditions of high food availability produce many more offspring that are not resistant to starvation (Gliwicz & Guisande, 1992), presumably because their offspring are unlikely to encounter significant food deprivation. In a similar kind of study, Kim & Thorp (2001) report that the solitary bee, Megachile apicalis, varies seasonally the allocation of resources to eggs. Spring females produce many small offspring, whereas summer females produce a few larger offspring. The production of large offspring later in the season is considered to increase the probability of overwintering survival. Recent work suggests that maternal effects acting upon the egg size/egg number trade-off may be adaptive primarily for the mother, increasing her fitness at the expense of offspring fitness (Einum & Fleming, 2000). Of course, parental care can extend well beyond the provisioning of eggs, and much of that care can be envisaged as a sort of protracted ‘packed lunch’. At its most simple, post-hatch care can include the provisioning of resources for larvae in the absence of contact with the parent (dinner is in the fridge, so to speak). However, the resources provided need not always be nutritional. For example, when females of the desert tenebrionid, Parastizopus armaticeps, are provided with supplementary food, they do not increase allocation to their young. Rather, they spend less time foraging and more time aiding males in digging natal burrows. The depth of natal burrows is related to larval survival because deep burrows reduce the probability of desiccation (Rasa, 1998). In this case, moisture rather than food represents the currency of the maternal effect, even though the availability of food for the mother initiates the effect. The longer-term consequences of parental provisioning can be dramatic. For example, larvae of the long-horned dung beetle Onthophagus taurus are provided with fragments of dung before the departure of their parents. In a fascinating example of a paternal effect, males that aid females in the provisioning process produce sons with horns, whereas males that do not help provision the nest produce males without horns. The extra dung provided by the helpful males is sufficient to induce horn production in their male offspring (Hunt & Simmons, 2000). The effects of such paternally determined variation in morphology for the population ecology of the beetles is unclear, but is very likely to influence the competitive ability of male beetles. In other words, maternal effects can influence the phenotype of offspring well beyond the first few days or weeks of life. In the outbreak insect, Epirrita autumnata, maternal environment still influences the consumption and growth of fifth-instar larvae and their subsequent pupal mass (Alonso et al., 2001). Given that pupal mass is a reliable indicator of fecundity in E. autumnata, maternal effects have a clear potential to influence population growth. The pupal mass of Papilio butterflies is also influenced by maternal diet (Thompson et al., 1990). Maternal effects can be of even longer duration. For example, egg-to-adult viability in Drosophila serrata is dependent upon both maternal and grandmaternal effects (Hercus & Hoffmann, 2000). Similarly, the ‘telescoping generations’ of many aphids, whereby the development of grandchildren commences within the grandmaternal body (Mousseau & Dingle, 1991), provides a substantial time-lag between the current environment and the expression of phenotype in offspring. Lest we consider grandparental effects as unique to insects, studies have shown that the mass of deer fawns and their susceptibility to wolf predators is a function of both maternal and grandmaternal environment (Mech et al., 1991). The life-history traits that are influenced by maternal effects are many and diverse (Mousseau & Dingle, 1991). For example, maternal effects have been invoked to explain variation in fly mating behaviour (Mangan, 1991), adult morphology (Bryant & Meffert, 1998), the choice of pupation site (Bauer & Sokolowski, 1988) and the competitive ability of offspring (McIntyre & Gooding, 2000). Maternal effects are the most important determinant of head width in alfalfa leafcutting bees, Megachile rotundata (Owen & McCorquodale, 1994) and a significant contributor to variation in the body size of adult carabid beetles (Desender, 1989). Maternal effects may be mediated by the interaction of mothers with natural enemies in the environment. For example, in the damselfly Coenagrion puella, a high ectoparasite load (Acari: Arrenurus cuspidator) on mothers stimulates the production of fewer, larger offspring (Rolff, 1999). Although the adaptive value, if any, of this maternal response is unclear, changes in offspring size and number have the potential to influence damselfly population dynamics. Within a single clutch of eggs, some female insects may allocate resources differentially depending upon the risk of predation. Female shield bugs, Elasmucha ferrugata, allocate more resources to individual eggs with low risk of predation than they do to eggs with a higher risk of predation (Mappes et al., 1997). In this case, female shield bugs guard their eggs from predators and are more likely to lose eggs from the edge of clutches than from the clutch interior (Mappes et al., 1997; Kudo, 2001). Central eggs are significantly larger than those at the periphery of the mass. Of course, maternal effects can also act to delay the response of individuals to immediate predation risk in the environment. For example, soldier production by the eusocial bamboo aphid, Pseudoregma bambucicola, cannot adjust immediately to an increase in predator abundance because the fate of developing eggs within the mother's body appears to be determined by prior environmental experience (Shibao, 1999). This time-lag in response to environmental conditions is precisely the kind of effect that can influence subsequent population fluctuations. The natural enemies of insects are themselves subject to maternal effects, and such effects have the potential to influence the population dynamics of both enemies and their prey. For example, the efficacy of the parasitoid Muscidifurax raptor declines rapidly in laboratory culture, in part because the fecundity of the parasitoid is largely determined by non-genetic maternal effects (Geden et al., 1992). Similarly, both development time and body mass of the pentatomid predators, Podisus maculiventris and Podisus nigrispinus, are influenced by maternal effects (Legaspi & O'Neil, 1994; Mohaghegh et al., 1998a,b). Dispersal, a key component of population change, can be influenced dramatically by maternal effects, and two pathways appear to be responsible. First, the environment experienced by the mother can influence the tendency of her offspring to disperse (Messina, 1993; Dingle & Winchell, 1997; Massot & Clobert, 1995; Diss et al., 1996). Second, maternal environment can affect the probability that the offspring will survive the dispersal process (Morse & Stephens, 1996). Examples of the former abound in the literature. Alate production by aphids is a classic case in which the environment of the mother (climate, density and food quality) influences the phenotype and dispersal ability of her offspring (Dixon, 1971; Mittler, 1973; Harrewijn, 1978; Messina, 1993). I doubt that many would argue that the production of winged adults and a sexual generation has no influence on aphid population dynamics, and this may be the clearest link between maternal effects and population change of which we are currently aware. Changes in dispersal ability have also been noted in other systems. For example, the tendency of gypsy moth larvae to disperse after hatching from their egg masses is related to the environmental experience of the mother (Diss et al., 1996). Likewise, the proportion of macropterous and micropterous offspring produced by some Japanese crickets is influenced by maternal effects (Masaki & Shimizu, 1995). Finally, maternal effects are a key component of phase change in locusts and contribute to the production of the migratory forms that are so devastating to developing agriculture (Islam et al., 1994). The influence of maternal environment on survival during dispersal is much harder to record and, as a consequence, fewer examples exist. However, variation in mortality during dispersal of the crab spider, Misumena vatia, is almost entirely dependent upon the foraging success and subsequent mass of the mother spider (Morse & Stephens, 1996). The study of Bruchid seed beetles has provided considerable insight into the expression of maternal effects. Maternal effects may influence adult dispersal (Messina, 1987), host preference and host expansion (Fox et al., 1997a; Messina & Slade, 1997; Fox & Savalli, 2000), egg size and viability (Fox, 1993; Fox et al., 1995, 1997b), larval survival (Fox et al., 1995; Fox & Savalli, 2000) and the development time of larvae (Fox, 1993; Fox et al., 1995, 1999b). Even ejaculate size of male beetles, a significant contributor to female fecundity, may be influenced by maternal effects (Savalli & Fox, 1998). Critically for population dynamics, the expression of maternal effects in seed beetles can be density dependent (Fox & Savalli, 2000), paving the way for time-lags in the action of density-dependent processes, and the potential induction of population cycles (Rossiter, 1991a, 1994). If you are a Bruchid beetle, chances are that your mother's environment has had a major impact on your life. Likewise, laboratory and field studies of crickets and grasshoppers have provided a wealth of information on the expression of maternal effects. Environmental variation experienced by mothers can influence wing size (Masaki & Shimizu, 1995), the morphology of sound-producing organs (Webb & Roff, 1992) and components of the calling song of males (Webb & Roff, 1992). Perhaps most significantly for population dynamics, maternal effects can have a strong influence on the induction of diapause and subsequent voltinism (Mousseau & Roff, 1989; Dingle et al., 1990; Mousseau, 1991; Groeters, 1994). For insects in general, rates of population growth can be affected markedly by the number of generations exhibited each year (Carrière et al., 1995; Hunter & McNeil, 1997). When diapause is facultative, the environment experienced by mothers can promote continued development when conditions are favourable, or induce diapause when conditions are unfavourable (Roff, 1983; Hunter & McNeil, 2000). Environmentally based maternal effects have been shown to influence diapause and/or voltinism in parasitic Hymenoptera (Brodeur & McNeil, 1989; Fabres & Reymonet, 1991), Lepidoptera (Sims, 1983a,b; Hunter & McNeil, 2000), Diptera (Saunders, 1987; Rockey et al., 1989, 1991; McWatters & Saunders, 1998; Webb & Denlinger, 1998) and Homoptera (MacKay & Wellington, 1977; Harrewijn, 1978). Simply put, when voltinism is variable for a given species of insect, maternal effects are often part of the story (Mousseau & Dingle, 1991). Despite Wellington's (1957, 1960, 1964) early interest, the expression of maternal effects in Lepidopteran pests has received less attention than it merits. However, there are convincing examples where maternal effects have been shown to influence a broad range of phenotypic traits related to population dynamics (Gould, 1988; Rossiter, 1991a,b, 1992, 1994; Rossiter et al., 1993; Diss et al., 1996; Behmer & Grebenok, 1998). For example, the allelochemistry of the parental diet can influence the performance of Heliothis virescens larvae, with the expression of the maternal effect dependent upon offspring environment (Gould, 1988). Essential nutrients (or the lack thereof) also invoke maternal effects. Phytosterols in parental diet, required for normal development of larval offspring, can have positive or negative effects upon the larvae of Plutella xylostella, depending upon whether the phytosterols are from host or non-host species (Behmer & Grebenok, 1998). Similarly, iron deficiency in the parental diet of the gypsy moth, Lymantria dispar, reduces the fitness of offspring regardless of the diet that they receive, and accounts for 28% of the phenotypic variance in larval performance (Keena et al., 1998). Indeed, environmentally based maternal effects have been shown repeatedly to influence the phenotypic traits of gypsy moth larvae and pupae, and provide an interesting example of the development of a theory of population dynamics from empirical studies of quantitative genetics and maternal effects. In 1990, following studies of the genetics and ecology of host use by the gypsy moth (Rossiter, 1987a,b; Rossiter et al., 1988), Rossiter demonstrated that the susceptibility of gypsy moth larvae to the bacterial insecticide, Bacillus thuringiensis, was determined in part by maternal provisioning of eggs (Rossiter et al., 1990). In subsequent studies of egg provisioning, Rossiter reported that daughters emerging from large gypsy moth eggs hatched earlier, grew faster and attained greater pupal weights and fecundities than did daughters emerging from small eggs (Rossiter, 1991b). Although egg size had a genetic component, there was also a strong influence of maternal diet and thus maternal effects contributed significantly to offspring development and fecundity. For example, mothers reared on damaged foliage of red oak, Quercus rubra, produced offspring with heavier pupae than did mothers reared on undamaged foliage (Rossiter, 1991a). In addition, the condensed tannin concentration of the maternal diet was related both to the length of the larval pre-feeding period and to the pupal mass of sons. When parents were fed on two different oak species, parental diet accounted for 24% of the variance in larval development time, whereas offspring diet accounted for 52%. These results led Rossiter (1991b, 1992) to suggest that maternal effects could play a hidden role in the population dynamics of the gypsy moth. Subsequent work by Rossiter developed along two lines; studies of the causes and consequences of maternal provisioning (Rossiter et al., 1993, 1996; Yerger & Rossiter, 1996), and the development of conceptual models of population dynamics that included time-lagged contributions of maternal effects (Rossiter, 1994, 1995, 1996, 1997, 1998). For example, the dominant storage protein in gypsy moth eggs is vitellogin. In experiments comparing quaking aspen, chestnut oak, red oak and pitch pine as parental diets, maternal diet was found to influence vitellogin levels in eggs (Rossiter et al., 1993). In addition, the solubility of the egg proteins was related to phenolic compounds in the eggs derived from the parental diet (Rossiter et al., 1996). In her conceptual models, Rossiter explored the consequences of time delays in the expression of environmental conditions for population stability, escape from regulation, and outbreak dynamics (Rossiter, 1994). Maternal effects have the potential to uncouple the density and quality of herbivore individuals from other factors in the environment, including host quality and the density of natural enemies. As with many time-lagged density-dependent factors, maternal effects can generate population fluctuations and cyclic dynamics (Berryman, 1999). Rossiter's work has stimulated both theoretical (Ginzburg & Tanneyhill, 1994, 1995; Berryman, 1995; Crone & Taylor, 1996; Ginzburg, 1998; Berryman & Chen, 1999; Benton et al., 2001) and empirical (Harrison, 1995; Myers et al., 1998; Erelli & Elkinton, 2000; Ergon et al., 2001) studies of the role of maternal effects in population dynamics. The theoretical work has largely centred around the appropriate model structure for incorporating maternal effects into population models and the relationships between model structure and the predictions that they generate (Ginzburg & Tanneyhill, 1994; Berryman, 1995; Benton et al., 2001). Even when authors disagree about the appropriate manner in which to model maternal effects, all agree that they have the potential to generate population cycles and may be difficult to untangle from other sources of delayed negative feedback (Berryman, 1995, 1999; Benton et al., 2001). Agreement has been less forthcoming in empirical studies of maternal effects in natural populations. Two field studies of the gypsy moth have reported effects of maternal environment on offspring performance (Diss et al., 1996; Yerger & Rossiter, 1996), whereas two other field studies have not (Myers et al., 1998; Erelli & Elkinton, 2000). However, in my opinion, all four studies are flawed. In each case, eggs or larvae were collected from different gypsy moth populations that varied in their population trajectories (e.g. population trough, population increase, population peak). Effects on larval dispersal, survival and/or growth were then compared to test the hypothesis that individuals from similar positions on the population trajectory would share phenotypic traits in common because of shared maternal experience. All four studies are therefore examples of ‘space-for-time’ replacements in which spatially disjunct populations at different growth stages are used to simulate changes in a single population over time. The problem is that maternal effects, by their very nature, are not amenable to space-for-time replacements. The studies do allow a test of whether or not populations at a particular phase of growth share common phenotypic traits. They do not allow a test of the hypothesis that changes in maternal provisioning within a single site over time are associated with temporal population dynamics. The critical question is whether females change their allocation to eggs over time as environmental conditions change, not whether those allocation patterns are in any way similar to those of other populations. In other words, there may not be a consistent pattern of phenotype among populations at the same stage of growth. However, each population may exhibit variation in allocation over time and time-lagged effects on performance sufficient to generate population cycles. As has been stated clearly in the past, the expression of maternal effects varies in space and there is no shortcut to exploring links between maternal effects and population dynamics in natural populations over time (Wellington, 1957; Labeyrie, 1988; Mousseau, 1991). Multigenerational studies of insect density, insect quality, and the environment (biotic and abiotic) are required for an adequate test of the maternal effects hypothesis of insect outbreak (Rossiter, 1994). The current state of affairs, then, is that there exists (a) strong empirical evidence from laboratory studies to suggest that maternal effects influence the life-history traits of insect pests, (b) recognition that maternal effects provide the potential for delayed negative feedback in populations, and (c) a well-developed theory linking delayed negative feedback to population dynamics, outbreaks and cycles. What we are still missing is clear evidence of population dynamics in the field that are driven by maternal effects. Although the work reported by Baltensweiler and by Nealis in Aberdeen is strongly indicative of maternal effects, demonstration of cause and effect remains elusive. As Berryman (1995, 1999) has pointed out, it will be difficult to untangle the influence of maternal effects from other delayed negative feedback processes such as wound-induced declines in plant quality or the influence of natural enemies. Maternal effects therefore fall squarely among the ranks of other ecological processes that must be explored with a combination of experimental studies, life table data and time-series analysis (Hunter et al., 1997; Hunter & Price 2000; Turchin & Berryman, 2000; Hunter, 2001). Examination of maternal effects in population dynamics have the added difficulty that ‘space-for-time’ replacements are inappropriate experimental techniques, and studies must be multigenerational within single localities. Whether maternal effects will prove to be major drivers of population change is not yet clear. However, it is clear that ignoring, or failing to control for, maternal effects may result in misleading interpretations of experimental and census data, whether or not such effects are a focus of the research (Mousseau & Dingle, 1991; Bernardo, 1996). I would like to thank the National Science Foundation (Grant DEB-9906366) for supporting our work on the population dynamics of insect herbivores. I would also like to thank my mother for her continued interference, genetic and otherwise, in my life. Accepted 5 November 2001