Introducing biological realism into the study of developmental plasticity in behaviour

Maternal effects

The earliest source of information in development is via the mother. She can change the composition of the egg before fertilization, either by changing resources for the embryo such as nutrients, immune factors or vitamins, or signals such as hormones. Fathers can play a role here, too. Their genes may affect which genes of the mother come to expression via genomic imprinting. In addition, the quality of the father can affect the female deposition of resources and signals. This has been extensively studied in birds in which the influences of a variety of environmental cues on egg composition and size have been demonstrated (for reviews see [17, 18]. Maternal effects on egg quality almost always end after oviposition in oviparous species although in fish species such as cichlids during parental care for the eggs urine containing parental hormones may enter the egg. This provides an opportunity for fathers too to affect egg composition. In placental animals there is a much longer and actually mutual exchange of substances between mother and offspring [19]. These prenatal parental effects are often overlooked and difficult to control for in studies manipulating postnatal environmental cures. Cross-fostering is sometimes claimed to control for maternal effects but disregards prenatal maternal effects.

One important topic associated with these maternal effects is the discussion whether mothers “actively or passively” bestow their eggs or embryos with resources and signals. In our opinion this is an inadequate dichotomy as “active” suggests a “decision” by the female (again a confusing word suggesting a conscious process) and passive suggests that the variation in maternal deposition is an inevitable consequence of environmental influences on her physiology. The real interesting issue here is to what extent mothers can regulate the deposition into eggs of resources and signals independent from what she would produce in her own circulation. This is discussed extensively in [20]. The current data suggest independent regulation as far as hormonal signals are concerned [21]. This would free the mother from a trade-off between allocation of these signals to her offspring and herself. This is likely to be different for resources such as antibodies and nutrients, unless these are available ad libitum. Clearly, for understanding functional consequences and trade-offs, we need to integrate proximate approaches with ultimate explanations.

Maternal and paternal effects can obviously occur in many ways also postnatally. An obvious pathway is via (pre- or) postnatal food provisioning that may provide the offspring a head start that has beneficial consequences throughout life, the so called ‘silver-spoon effect’ . In species in which offspring size or body mass is related to mother's size or body mass, this may translate in transgenerational effects (e.g. [22]). Such effects are also conceivable due to habitat imprinting, in which offspring choose their habitat to reproduce depending on what they have experienced when they were young as a consequence of parental nest site choice. Interestingly, such transgenerational effects are easy to misinterpret as genomic heritability but they are in fact non-genomic. These effects might be caused by epigenetic processes in which DNA expression is affected by environmental input, for example, by methylation or his tone modification that has to occur every generation again [23]. Many other maternal effects can be envisaged, ranging from food preferences, immune defences, and migration routes to partner preferences and tool use.

Direct environment effects

The above discussed maternal or parental effects are often labelled as indirect environmental effects as the effect of the environment on the offspring is transmitted through the parents. Direct environmental effects have received more attention in the previous century (e.g. [24]) or have not been recognized as indirect environmental effects. In especially precocial species many environmental influences can be direct, although by determining the habitat in which the offspring experience these influences parental effects can be still playing a role. For example the level of early aggressive interactions between black-headed gull chicks affect their early endogenous testosterone production having long-term consequences for their sensitivity to the hormone later in life [25]. This early social experience is due to nest site choice of the parents so actually a parental effect. Also the social environment experienced early in life, such as social group size in cooperatively breeding species, is usually considered as direct environmental influence with potential strong effects on future social performance [26], but group size itself can be a function of parental quality. These examples suggest that environmental factors can constitute both direct and parental effects.

Similar to indirect (parental) effects, direct environmental effects can be induced by all possible features of the early environment including habitat (e.g., [27], food quality (e.g., [28], [29]) and availability or social context (e.g., [30]. These early, direct experiences often result in long-term preferences for similar features later in life as has been shown, for example, in experiments testing for habitat imprinting ([31]. Quality differences of the early environment such as early food conditions can exert long-term effects on fitness though carry-over effects [29, 32], which work analogously to silver-spoon effects caused by parental provisioning (see above). Long-term effects induced by direct early experience can become manifest through stable reprogramming of gene expression profiles [33], which has been hypothesized to be caused by epigenetic modifications such as DNA methylation or chromatin changes.

Interactions between early (in)direct environmental effects and later experience

Parental effects and effects directly induced by environmental conditions experienced early in life can, in our opinion, in many cases not be interpreted when the environmental effects on the offspring later in life are not taken into account. Below we sketch six scenarios in which the later experience can interact with early experience.

Support from theory

Models on the evolution of adaptive plasticity generally agree on that plasticity requires at least three ingredients. A phase where organisms are sensitive to environmental cues that predict future conditions; a period where a plastic response is shown to match the forecasted conditions; and, finally, a phase when the produced phenotype is exposed to the selective environment that determines which phenotype has the highest fitness. Most models on adaptive plasticity allow these three ingredients to occur only once in a life history (reviewed in [55]). From empirical studies we know that this view is a gross oversimplification of the process of plastic development (e.g., [10, 56]). Plastic adaptations to one state of the environment can be halted or even fully reversed when later in life individuals perceive cues signalling that environment has switched to the opposite state [26], and some organisms continue sampling and adjusting their phenotype to perceived environmental cues well beyond sexual maturity (e.g. [57] or even lifelong [58]). Thus these kinds of dynamics of plastic phenotypic development need to be taken into account in theoretical models if they should aid our understanding of developmental plasticity. The key conclusions of theoretical studies incorporating this complexity are summarized below.

In a first step towards modelling the dynamics of plastic development, reversible plasticity was considered [59]. Using predator-induced structures as example, a plastic response was modelled that allowed previously induced defence structures to be reduced again when predation risk vanished. The model specifically investigated the role of phenotypic response lags relative to the rate of environmental change. Such response lags are thought to help avoid overly quick responses towards incomplete or unreliable environmental information and thereby to save costs of plasticity. If response lags were short or the environmental information reliable reversible phenotypic plasticity readily evolved and the phenotypes closely tracked the state of the environment [59]. Time lags can be costly themselves, however, if they prevent an adequate response to a new environmental risk, e.g., the presence of a new predator. The model indeed predicts that if environmental information is incomplete and time lags are long the optimal strategy is to converge towards a generalist strategy that can cope with a broad range of environmental states, reducing plasticity.

The next step towards a more realistic view on development was taken by [55], who modelled development as a ‘constructive process’, where individuals adapt incrementally to their local environment. The authors explicitly model two strategies to be linked by a trade-off: (i) response lags, during which information about the environment is collected, or (ii) phenotypic specialization towards the current environmental state. At each of a defined number of time points, model organisms were allowed to choose between either promoting phenotypic specialization or sampling the environment without phenotypic change. The optimal strategy depends on the reliability of environmental cues. Spending more time specializing may in the end lead to a better adapted phenotype, but only if the initial estimate on the environmental state was correct. Spending more time sampling reduces the uncertainty about the state of the environment and may therefore avoid maladaptation. Longer sampling periods emerged when cues were moderately informative, but not when they were highly or weakly informative.

Finally, in a recent theoretical study optimal plasticity trajectories were analysed throughout the lifetime with organisms having full flexibility to repeatedly adjust their phenotype to one environment or reverse these specialisations to adapt to an alternative environment [60]. Organisms can sample and track the environmental states at different ages until death. All optimal reaction norms of phenotypic plasticity turned out to have a characteristic shape, which entails a broad region where after an environmental switch no or little phenotypic adjustments are made. This “plateau” of the reaction norm causes time lags of plastic responses as an emergent feature of the model. This suggests that response lags in phenotypic adjustment can be part of an optimal strategy rather than being caused by constraints. These response lags only vanish when plasticity costs are zero. Several common patterns were observed in a majority of model environments. (i) Typically there was an early-life peak of phenotypic adjustments once young individuals had accumulated sufficient environmental information. (ii) Then, after a period of reduced plasticity, a second, broader peak of plasticity follows in response to environmental changes during life time. (iii) Plastic adjustments at the end of life were rare, because the costs for such adjustments have to be paid, while the benefits are unlikely to be reaped before death. These results may at least partly explain the existence of multiple sensitive windows observed in empirical studies.

In summary, the reviewed models suggest that selection does not only favour an early single sensitive window with organizing effects, but also reversible plasticity under certain conditions, lagged responses to environmental change and multiple windows of enhanced plasticity during lifetime. The latter requires extended periods of time when organisms are sensitive to environmental cues. Extended sensitivity in combination with lagged responses open the possibility that organisms collect environmental information during multiple periods in life before they show a phenotypic response that integrates the different information they had collected. In the following sections we examine the ways how animals integrate information from different life stages when mounting plastic responses during development.

Interactions between mother and offspring

In almost all research on prenatal maternal effects it is assumed that the embryo is just a slave of the mother in responding to maternal provisioning. In the human literature it is, however, well established that there is mutual exchange of substances such as hormones between mother and embryo affecting and manipulating each other in their own interest [19]. This substantially complicates the study of development. This is much easier in oviparous species in which the scope for interaction between embryo and mother is very limited after oviposition and can further be reduced by artificial incubation. There is recent evidence that in such species the embryo plays an important role, too. For example, embryos of turtle, fish and bird species are surprisingly capable of converting the maternal hormones to biological inactive or active components [61–63]. Moreover, prenatal influences such as incubation temperature affect the embryo's own hormone production. For example, incubation temperature affects hormonally guided sexual differentiation in the leopard gecko [64]. In wood ducks, incubation temperature affects the chick's own production of thyroid hormones, involved in growth and metabolism [65]. This begs the intriguing question to what extent the embryo might or might not “listen” to the maternal signals depending on its context [17]. Obviously, maternal effects may be so important because only the mother may have information about the best developmental pathway at this stage, but cues related to egg quality, incubation temperature, or vocalisations by other embryos might provide relevant cues for the embryo to determine what to do with the maternal signal.

Also after the embryonic phase direct experience of offspring can abolish or reverse maternal influences on behaviour. In the cooperatively breeding cichlid Neolamprologus pulcher dominant females produce smaller eggs in groups with many helpers compared to females in small groups [66]. Offspring born in small or large groups but reared separately from parents and helpers differed in their tendencies to show submissive and aggressive behaviour later in life. In contrast, siblings of these offspring reared for two months within their natal group also differed in later social behaviour, but showed exactly the opposite behavioural tendencies, suggesting that the direct social experience had fully reversed initial, maternal influences on behaviour [26].

Interactions between early and late environment

As predicted by the model of [60], animals collect information from the environment repeatedly during multiple stages in life and use this additional information to correct or modify initially pursued developmental trajectories. Tracking of environmental change and reversal of early specialization might be expected to occur particularly in behavioural phenotypes as remodelling of behaviour is assumed to imply less plasticity costs than, for instance, morphological structures. Indeed there are examples where behavioural differences induced by early social experience are fully abolished if opposite experiences are made in the sub-adult stage. When laboratory rat pups experienced low levels of maternal care they expressed low-level maternal care as adults themselves. The effects of poor maternal social stimulation were fully abolished, however, by housing rats in socially enriched environments after weaning [67]. What looks like a reversal on the behavioural level may however also be a compensatory effect at the brain gene expression level [68]. The most prominent and efficient mechanism of information updating is learning. Guppy males (Poecilia reticulata) reared in visual contact with conspecific females developed a strong tendency to perform forced copulations, whereas those reared with visual contact to males developed overly long poor courtship displays. Both of these exaggerated behavioural tendencies were reduced to a ‘normal’ level after only 2 days of sexual experience as adults in direct contact with conspecifics [69].

Given the accumulating theoretical and empirical evidence that animals can collect and integrate environmental information across multiple life stages, experiments on development should now move a step forward towards more complex designs addressing explicitly the integrative abilities of developing organisms by including more than two developmental stages. A recent study is a promising example of such a more complex design [32]. In order to test the match-mismatch hypothesis, nutrition was manipulated at two levels at the larval and at the nymphal stage of burying beetles (Nicrophorus vespilloides) in a crossed design and the joint effects of these treatments on intra-sexual competitiveness were investigated in adult beetles again at two levels of competition. Larval and nymphal nutrition jointly affected male fighting success through indirect (larval) and direct (nymphal) silver-spoon effects on body size and fighting ability, respectively.

Interactions between different maternal effects

Mothers provide their offspring with a wide array of resources and signals and the effect of one of these on the phenotype of the offspring might depend on others. For example, maternal stress in mammals and birds has an effect both on embryonic cortisol exposure as well as on blood and nutrient provisioning, making it hard to determine the relevant pathway explaining the effects on development [79, 80]. In birds, mothers provide their eggs among many other things with immune factors, carotenoids and androgens that mothers may or may not adjust to each other [81]. It has been repeatedly demonstrated that testosterone in the yolk of bird eggs can be immunosuppressive for the embryo (for a review see [18]. It has been suggested that the effect of elevated androgen concentrations in the egg may be beneficial in good quality eggs but detrimental in poor quality eggs [17, 82]. The same holds for eggs sired by good and poor fathers, [83]. Detrimental effects of yolk androgens may be counteracted by enhanced provisioning of maternal antibodies and carotenoids. Indeed, several studies have analysed correlations among different bird egg components, finding correlations that can differ both within populations, among females, and among populations. This suggests that mothers may be able to uncouple these components in an adaptive way (for a review see [76]. This obviously poses a challenge for experimental studies in which normally only one component is measured and manipulated and warrants many more experimental studies in which more than one component is manipulated.

Interactions between different direct experiences

Generally, multiple environmental factors can have independent, additive or interactive effects on the development of phenotypic traits. Independent effects of the manipulated factors (Fig. 3a) occur quite often. Within the same experiment, different factors can independently affect different aspects of behavioural tendencies. For example, rats were reared in a balanced design either in isolation or in same-sex conspecifics groups and either in a barren or in an enriched cage. Isolation-reared rats had enhanced general activity as adults, irrespective of the physical enrichments during rearing, whereas, early enrichment improved habituation responses and spatial learning abilities of adults regardless of their social rearing environment [84]. The adaptive relevance of these results is unclear, however, as the rats may not have expressed evolved responses towards the highly artificial laboratory rearing conditions.

If we want to study evolved reaction norms in the laboratory, we should carefully think about the most important natural environmental factors for our study species, and vary them within the natural parameter range of these factors. For many animals two very important selective factors are food availability and predation risk, which can be connected through a trade-off limited by a common time constraint. Whether this potential trade-off is shaped by early experience has been studied in guppies (Poecilia reticulata) reared at two levels of early-life food abundance crossed with two levels of perceived predation threat [85]. Interestingly, anti-predator behaviour was only influenced by early perceived predation risk. Similarly, when juvenile cooperatively breeding cichlids (N. pulcher) were reared at two levels of social complexity crossed with two levels of perceived predation risk, predator avoidance behaviour later in life was only influenced by early predation threat [86]. The same early-life influences can trigger quite divergent long-term responses in males and females [87], however. Crickets (Telogryllus commondus) were kept as juveniles at two density levels crossed with three levels of male calling intensity. In females, the latency to respond to prospective mates was only influenced by the early calling environment. In contrast, male life span depended only on rearing density, whereas male age-specific calling effort was influenced by both factors interactively.

Growing evidence suggests that under certain conditions organisms may need joint information about multiple environmental factors in order to express adequate plastic responses during development [88]. For example, Daphnia magna required combined cues on current resource availability, photoperiod and population density in order to induce a switch from clonal reproduction to the production of sexual, dormant eggs, which can endure harsh conditions [77]. Additive effects of two factors become apparent by parallel reaction norms, but they can be interpreted in two ways; a second factor can be viewed to enhance the effect induced by a first factor (Fig. 3b, upper panel) or to reduce its effect (Fig. 3b, lower panel). The latter case may indicate that a trade-off is involved between the responses to both factors. An example of an enhancing effect is that the presence of early predator cues enhanced growth and thus final size in both low-food and high-food reared guppies [85]. An example of a trade-off is that anti-predator responses may have to be traded off against the development of morphological structures. In the absence of predation risk, wood frog tadpoles (Rana sylvatica) develop increasingly longer guts with increasing conspecific competition [89]. The same positive relationship between gut length and competitor density was found if tadpoles were reared in different concentrations of predator cues, but the higher the perceived predation risk the shorter the gut length was across all density treatments. Instead, under predation risk tadpoles allocated resources more into tail development.

Certainly the most interesting outcomes of designs with multiple environmental factors are interactive effects where reaction norms differ in slope. Most striking are examples where the slopes of reaction norms have opposite signs (Fig. 3c, upper panel). This was found, for example, when the influences of early social and predator experiences on the social behaviour of the cooperatively breeding cichlid N. pulcher were investigated. Different behavioural specializations in the social domain of the four treatment groups indicated that these fish required the information of both early environmental experiences to develop adequate behaviours [86], a fact that was overlooked when only the early social environment had been manipulated [28, 76, 90].

Unless full factorial experiments are employed, environmental effects may be entirely overlooked because they can be masked by a second factor. This can happen when one reaction norm has a slope different from zero and the other one is flat (Fig. 3c, lower panel). This has been demonstrated in crickets (Telogyllus commodus), in which a protein reduced diet resulted in the development of smaller adult size. These size differences vanished, however, when crickets were provided with acoustic cues signalling high future intra-sexual competition. By anticipating future competition, crickets enhanced growth during their last nymphal stage and were thereby able to compensate fully for initial grow deficits [91]. Interestingly, also the absence of an environmental trigger can have masking effects. Tail morphology of tadpoles is a textbook example of predator induced plasticity. When the population density of tree frog tadpoles (Hyla femoralis) was varied, no plastic response in tail morphology was observed, although low-densities are an indicator of higher potential predation risk. As soon as chemical predation cues were added, however, tail morphology readily varied with population density as predicted [92].