Introduction

Here, we provide an overview of studies investigating the impact of early life experiences on the development of individual differences in neuroendocrine stress responsiveness in adulthood. Evidence for this phenomenon is largely provided by research on developmental programming of the stress system. This field of research investigates the mechanisms underlying the impact of environmental stressors or exogenous glucocorticoids during critical periods of development on sensitivity to stress and vulnerability to stress-related diseases throughout the lifespan.

The term “developmental programming” derives from the concept of “developmental origin of adult diseases” introduced more than 20 years ago by Barker and colleagues and was based on a large body of epidemiological research documenting the relationship between low birth weight and an increased risk of developing metabolic and cardiovascular disorders (Barker 1992a, b, c). These findings led to what is currently known as the Barker hypothesis (Barker 1997). This concept has been since extended to include “developmental programming of the hypothalamic–pituitary–adrenal (HPA) axis” (Andrews and Matthews 2004; de Kloet et al. 2005; Matthews 2002; Meaney et al. 2007; Seckl 2008). This is achieved through the actions of specific environmental cues acting at a specific time during development, and which result in permanent alterations in the functioning of the HPA axis (Andrews and Matthews 2004; de Kloet et al. 2005; Matthews 2002; Meaney et al. 2007; Oitzl et al. 2010; Seckl 2008). Why is this relevant at all? In addressing this question, preclinical and clinical evidence suggests that this remarkable phenomenon has relevance for the etiology of some of the major mental disorders triggered by stressful life events, including depression and posttraumatic stress disorder (Heim et al. 2008; Heim et al. 2004; Meaney et al. 2007; Seckl 2008).

Traditionally, this research has focused on detrimental consequences of stress but far less on the ability to develop resilience to stress or stress-related diseases. Recent findings are challenging this view and suggest that the outcome of early experience is not necessarily deterministic nor can it be perceived as good or bad, in absolute term. Several environmental, cognitive, and genetic inputs are involved in an intricate manner in the modulation of the outcome of early experiences. We first address instances where environmental influences in early life exert powerful effects on later stress phenotypes, even to the extent of overriding an individual's genetic attributes. Secondly, we address the evidence showing the crucial role of particular genetic constituents in modulating the outcome of these environmental effects. Thirdly, we address the increasingly recognized role of nonshared environmental experiences that are unique to the individual, in the outcome of gene × environment interplays.

Environmental influences

Animal models: postnatal manipulations

The field of developmental programming has a long history of studies on the importance of early experiences for the development and establishment of individual phenotypes (Levine 1957). This led to the emergence of a line of studies in rodents, primates, and humans until today. Broadly, these studies can be divided into investigating either prenatal or postnatal manipulations. Here, we will focus on the impact of several postnatal paradigms. For a review on the impact of prenatal manipulations, we refer to previous literature (Glover et al. 2009; Kinsella and Monk 2009; Lupien et al. 2009; Maccari and Morley-Fletcher 2007).

Early handling

In the 1950s, the late Seymour Levine made a fascinating discovery (Levine 1957). He exposed neonate rat pups to daily handling sessions, which consisted of brief maternal separation periods (<15 min) between postnatal day (pnd) 1 and 21. The outcome of these manipulations was surprising and unexpected at the time. Levine, and others, found that early handling (EH) induced long-lasting changes in adult phenotype such as HPA axis hyporesponsiveness (Levine 2005; Meaney et al. 1985; Meerlo et al. 1999), reduced emotionality (Meerlo et al. 1999), and increased cognitive performance (Kosten et al. 2007) in EH animals as compared to rats raised in undisturbed laboratory conditions, i.e., nonhandled (NH). However, the use of such control groups was recognized to be problematic later; see reviews (Macri et al. 2004; Macri and Wurbel 2006; Pryce et al. 2005). Because the EH procedure was considered at that time to be a stressful experience, these findings challenged the dominant theory stating that early life stress invariably contributes to the development of “emotional instability”. Instead, the findings from Levine demonstrated that, in some instances (e.g., via EH), exposure to “moderate stress” in early life appeared to be beneficial for the infant by promoting a greater ability of the organism to adapt to psychological and physiological stressors in adulthood (Levine 1957). This same principle also serves as the basis for the stress inoculation-induced resilience theory developed several years later (Levine and Mody 2003; Lyons et al. 1998; Lyons et al. 1999).

Maternal separation

Over the years, new paradigms were introduced in an attempt to also study the mechanisms underlying developmental programming following exposure to more “adverse” experiences (Plotsky and Meaney 1993). Maternal separation (MS) consists of prolonged periods of maternal absence ranging from 1 to 24 h. The reported effects of MS appear to be more controversial compared to the effects of EH, in part because of the substantial variety in different experimental procedures across different laboratories in terms of duration, frequency, age of onset of the separation, and the choice of control group (Lehmann and Feldon 2000; Lehmann et al. 2002; Rosenfeld et al. 1992). Nevertheless, MS appeared to “program” the functioning of the HPA axis. As expected, this manipulation was reported to yield a more severe outcome, opposing the effects of EH, including HPA hyperresponsiveness following stress (Plotsky and Meaney 1993), increased emotionality (Kalinichev et al. 2002b), altered sensitivity to opioids (Kalinichev et al. 2002a), and impaired cognitive performance (Levine 2005). For an extensive review of early life models, see Levine (2005) and Schmidt (2010).

Maternal mediation hypothesis

The use of the EH model raised an important question: How can short episodes of maternal absence result in such profound and enduring effects on adult stress phenotype? The “maternal mediation hypothesis” was proposed for the first time as part of the mechanism underlying the lasting effects of EH by Smotherman and Bell (1980). This theory postulates that the outcome of postnatal manipulations (such as EH, MS, and NH) is mediated by changes in maternal behavior directed towards the offspring upon reunion after a given period of mother–infant separation (Denenberg 1999). It was observed that brief (15 min) episodes of EH resulted in increased active maternal care as shown during observation periods over the entire day. Longer periods (4 h) of MS yielded an increase in active maternal care only directly after reunion of the dam with the pups but not at any other time point, leading overall to differences in the amount/quality of maternal care received by EH versus MS pups (Pryce et al. 2001). This suggests that the amount and quality of maternal care, at least in part, mediate effects of EH and MS on functioning of the HPA axis in the offspring.

However, certain findings challenged this theory. Although these studies clearly demonstrate the importance of maternal care for epigenetic programming of gene expression and thus provide support in favor of the maternal mediation hypothesis, for instance, Macri and coworkers (Macri et al. 2004) reported inconsistencies in the maternal mediation hypothesis. They showed an overall increase in maternal care following both EH and MS. Their findings revealed that following MS dams increase their care to such an extent that they fully compensate for the separation time and reach a level comparable to EH dams. Since EH and MS offspring display significantly different endocrine and behavioral stress responses in later life, it was concluded that maternal care cannot be the only mediator driving the effects of the postnatal manipulations (Macri et al. 2004; Macri and Wurbel 2006).

Tang and colleagues (Tang et al. 2006) have also tested the maternal mediation hypothesis in an original study. Using a paradigm in which half of the litter is stimulated by exposure to a novel environment, whereas the other half stays in the nest, Tang and colleagues have dissociated between the two potential sources of contribution to the development of the neonate—one source being the indirect effect of maternal care toward the stimulated (novelty-exposed) versus the home-staying offspring, and the other source being the direct effect of stimulating the pups through novelty exposure. It was reported that the novelty-exposed pups showed functional enhancement in term of cognition and endocrine responsiveness. However, this enhancement occurred in the absence of increased maternal care toward the novelty-exposed pups. Therefore, these findings suggest a revised maternal modulation hypothesis stating that (1) nonshared experiences, unique to the individual, have a direct effect on the offspring, potentially leading to alterations in adult phenotype, and (2) maternal behaviors can attenuate or enhance these effects, modulating the enduring effects on phenotype (Tang et al. 2006). These findings clearly indicate the importance of including the study of nonshared factors in addition to the study of shared factors within a litter, the later often being the only variable studied in these paradigms. This issue will be addressed into more detail later in this review (see Nonshared environmental influences section).

Naturally occurring variation in maternal care

The most compelling set of evidence on the importance of the amount and quality of maternal care on the development of the stress-regulating system came from the elegant work of Meaney and colleagues (Liu et al. 1997). Employing a noninvasive naturalistic approach, they studied the impact of naturally occurring variation in maternal care on the development of the HPA axis in rodents. This model is based on extreme differences among lactating rats in the frequency of licking and grooming (LG) they provide to their pups. They show that variation in the amount of maternal LG, a form of tactile stimulation, modulates the development of the structure and function of the neural circuitry underlying stress regulation, emotionality, and cognitive processes (Bagot et al. 2009; Caldji et al. 1998; Champagne et al. 2008; Liu et al. 2000; Liu et al. 1997; Smit-Rigter et al. 2009). Reminiscent of the outcome of EH, offspring of high, relative to low LG dams, show decreased behavioral and endocrine responsiveness to stress, reduced emotionality, and enhanced performance in tests of spatial learning and object recognition (Caldji et al. 1998; Liu et al. 2000; Liu et al. 1997). These effects are largely reversed with cross-fostering, in which the biological offspring of a high LG mother is cross-fostered to a low LG mother or vice versa. This suggests that variation in maternal care transfers phenotypic differences to the offspring in a nongenetic way, providing a clear demonstration of an environmental effect mediated by maternal care (Francis et al. 1999).

It should be noted that the early environment not solely consists of maternal factors. Other environmental factors also contribute in shaping later life outcome by affecting developmental trajectories. Several studies have indicated the importance of peer interactions. For instance, it has been reported that monkeys deprived of peer contact show abnormal development in terms of social behavior and emotional responses (Harlow 1969). Additionally, Branchi and colleagues, using a rodent model for social enrichment, have attempted to disentangle the effects of maternal versus peer interactions. Their findings suggest an important role for peer interaction in shaping the social and emotional phenotype in later life (Branchi 2009).

The stress-hyporesponsive period

The outcome of early life experiences largely depends on the timing, frequency, and duration an individual is exposed to particular environmental experiences (Champagne et al. 2009; Enthoven et al. 2008a; Meaney and Aitken 1985; van Oers et al. 1998a). For instance, the impact of EH procedures on adult phenotype appears to be more profound if they are performed during the early postnatal period as compared to later in the postnatal period (Meaney and Aitken 1985). This is important since the early postnatal period coincides with onset of the stress-hyporesponsive period (SHRP). The SHRP begins in the first postnatal days and terminates around pnd 14 in rodents (Levine 2005; Schmidt 2010). During this period, the neonate's adrenals are hyporesponsive to mild stressors (e.g., novelty exposure) that are capable to trigger a profound glucocorticoid response in the adult. While during the SHRP the adrenals are hyporesponsive, the brain does respond to novelty and other type of stressors. Maternal presence via active maternal care and feeding is suggested to actively regulate the responsiveness of the neonate's HPA axis during the SHRP (Levine 1994; Suchecki et al. 1993). Therefore, maternal presence serves to “buffer” the impact of stressors on neonates.

MS episodes will exert stronger impact especially if they occur within rather than outside the SHRP. The outcome of MS can be reversed by mimicking and reinstating aspects of maternal behavior during the separation procedure. For instance, MS-induced increase in glucocorticoid secretion was prevented by reinstating feeding via a cannula connected to the stomach, while artificial tactile stimulation gently administered with a paint brush during the separation period was required to restore pituitary adrenocorticotropic hormone release and c-fos mRNA expression in the paraventricular hypothalamus (van Oers et al. 1998b).

Recent findings from our laboratory revealed another issue to take into consideration when designing a study including the use of postnatal manipulations. Such issue concerns the impact of repeated MS during the SHRP. In a series of experiments, Enthoven and colleagues reported that neonates, as expected, responded to 8-h MS with a slow increase in HPA axis activation reaching a significant increase in glucocorticoids after 4 h. Interestingly, if the 8-h MS was daily repeated, the MS-induced HPA axis response became readily desensitized. From a psychological perspective, these findings suggest that pups may be able to predict the return of the dam and the reinstatement of maternal care, hence preventing the MS-induced HPA activation (Enthoven et al. 2008b).

It is also of significant relevance where the pup experiences repeated MS. We showed that repeated MS in the home environment result in habituation of the HPA axis activity in response to maternal absence. Remarkably, we observed that these pups continue to respond to an additional 30-min exposure to a novel environment. We propose that after repeated MS episodes, the neonate's HPA axis stays on alert and will be activated only when the return of the dam cannot be predicted (Enthoven et al. 2008b). Specifically, under conditions of repeated home MS, the adrenals become hyperresponsive to novelty. However, if the 8-h MS procedure is performed by placing the pup repeatedly in a novel rather than the home nest, the adrenals becomes hyporesponsive to novelty stress (Daskalakis et al., unpublished observations).

Another important aspect is the state of maturity of the neonate's nervous system. The developing brain is extremely sensitive to all types of sensory input such as those incurred through maternal LG. This is primarily due to the abundant proliferation of synaptic connections in the postnatal periods. Timing is therefore crucial in the newborn's brain. During the SHRP, the developing brain is able to integrate sensory input into circuits that control, for example, attachment to the caregiver (Moriceau and Sullivan 2004), or fear-motivated behaviors, and stress-induced HPA axis activation (Wolke 1987). Therefore, the maturity of the sensory system during the SHRP will also influence the degree by which glucocorticoids, environmental cues (e.g., maternal LG), or stress (e.g., EH and MS) will alter the development of these circuits.

The mechanisms underlying the ability of pups to make predictions about the environment are unclear. One possibility involves aspect of the neonate's circuitry (locus coeruleus and olfactory cortex) dedicated to odor learning (Moriceau and Sullivan 2004). Since neonate rodents depend on their mother for survival, this system facilitates the attachment to the caregiver. During the SHRP, when glucocorticoid levels are usually low, the neonate shows an increased ability to form odor preferences to maternal and other odors, and a decreased ability to form odor aversions, even to negative stimuli. This is believed to facilitate the establishment of a secure relationship with the caregiver (Moriceau and Sullivan 2006). When glucocorticoid levels increase and SHRP is disrupted, for instance due to MS, the amygdala gets activated, and a switch to odor aversion is observed (Moriceau et al. 2009; Raineki et al. 2010).

From these findings, a number of inferences can be made. Firstly, the predictability of the manipulation and the context in which they occur (novel vs. home, as also indicated by Tang and colleagues (Tang et al. 2006)) play a role in the outcome of the manipulation. This knowledge contributes to the understanding of the substantial variation in outcome of postnatal manipulations such as EH and MS as reported above in this review. Secondly, interpretation of the findings gathered from several animal models ranging from rodents and guinea pigs to nonhuman primates, which are born with differential degrees of brain maturity at birth, needs to be analyzed carefully before any comparison with humans can be made (Kapoor et al. 2006).

Epigenetic programming of the HPA axis

It was reported that changes in HPA axis activity and hippocampus-specific changes in GR expression levels, which are a prominent outcome of the EH paradigm, also occurred in response to naturally occurring variation in maternal care. Differential glucocorticoid responsiveness causing changes in synaptic plasticity is believed to underlie the vast array of neuroendocrine (Liu et al. 1997), behavioral (Caldji et al. 1998; Menard and Hakvoort 2007; Zhang et al. 2005), and cognitive alterations (Bagot et al. 2009; Champagne et al. 2008) observed in these early life stress paradigms (Levine 2005; Szyf et al. 2007).

However, a major breakthrough in the history of the field of developmental programming came with the discovery of epigenetic modifications in the promoter area of the GR gene, revealing a mechanism underlying these environmentally driven effects on later life stress phenotype. It was shown that increased levels of maternal LG during the first week of life alter the methylation pattern of the GR gene in the hippocampus of the offspring (Weaver et al. 2004). These changes persist into adulthood and alter the expression of the GR throughout life via modification of the chromatin structure. Cross-fostering of the offspring shows a complete reversal of methylation patterns, demonstrating that DNA can be structurally modified (without alterations to sequence) through environmental influences, thus leading to changes in gene expression (Meaney and Szyf 2005; Weaver et al. 2004).

The significance of these findings in the field of psychiatry is unclear, but recent studies in humans revealed that epigenetic programming of the HPA axis via changes in DNA methylation of GR may occur in human infants born to mothers who experienced depression during pregnancy (Oberlander et al. 2008). Additionally, there are indications of epigenetic regulation of GR in the brains of individuals with a history of adverse childhood experiences who committed suicide following a stressful life event (McGowan et al. 2009).

Evidence for developmental programming in humans

For obvious reasons, human studies on the programming effects of early life experiences are less numerous and more complex compared to the previously described animal studies. Since it would be unethical to expose the developing infant to the experimental manipulations that are imposed on the animals, advancement in this field relies on naturally occurring events and retrospective reports.

A series of studies from the groups of Heim and colleagues as well as Rinne and colleagues have convincingly shown the impact of early adverse events, such as childhood abuse, on alterations in the functioning of the HPA axis, as a consequence leading to an increased risk to develop psychiatric dysfunctions (Heim et al. 2000; Heim et al. 2008; Heim et al. 2004; Rinne et al. 2002). A dose–response relationship has been described between the number of experienced childhood adversities and mental health score in later life (i.e., probability of lifetime depressive disorders; Chapman et al. 2004; Edwards et al. 2003).

Besides severe forms of adversity such as emotional neglect accompanied by sexual or physical abuse, there is also evidence that milder forms of adversity are associated with increased risk for stress related-pathologies. For instance, not only early life socioeconomic disadvantage (Wickrama et al. 2005) but also subtle differences in parenting style (Feinberg et al. 2007; Feinberg et al. 2000; Reiss et al. 1995) appear to affect health status of the individual in later life.

Therefore, there is no doubt that early adversity plays a crucial role in programming the development of a range of physical and psychiatric disorders. As it is clearly shown in animal studies, vulnerability to diseases is likely to be mediated (at least partially) via the effects of early adversity on the functioning of the HPA axis. Several studies show the association between early life adversity and enduring sensitization of the responsiveness of the HPA axis in humans. For instance, alterations in basal as well as stress-induced HPA axis activity at different life stages have been reported in human subjects exposed to adversity in early life (Gutteling et al. 2005; Heim et al. 2000; O’Connor et al. 2005; Rinne et al. 2002).

Maternal influences in humans

As with animal models, it has also been demonstrated in humans that the mother (or another caregiver) plays a mediating role in the regulation of the HPA axis in children (Gunnar and Donzella 2002). Several studies show that when children are exposed to adequate care, they display diminished cortisol responsiveness, increased threshold to evoke a cortisol response to various stressors (Gunnar and Donzella 2002), and a better cortisol recovery after stress (i.e., glucocorticoid negative feedback; Albers et al. 2008). This is explained by suggesting that children, under high care-giving conditions, anticipate that a caregiver will protect them, and therefore, they feel able to cope with a threatening situation (Gunnar and Donzella 2002). This phenomenon could be an analog of the rodent SHRP, although the time window for this period is less well defined in humans (Gunnar and Donzella 2002). More lines of evidence for maternal mediation come from a study showing that prenatal stress is associated with a reduction in hippocampal volume only in combination with low levels of postnatal maternal care. This suggests that the impact of prenatal experiences is mediated by the postnatal maternal environment (Buss et al. 2007).

The role of genetic variation in the outcome of environmental influences

Genetic modulation in humans

Recent evidence has revealed that despite strong environmental influences, not every individual exposed to adverse experiences (up to 50% of cases) develops a stress-related psychiatric illness in later life (Bonanno and Mancini 2008; Caspi and Moffitt 2006; Rutter 2009, 2010). Clearly, the degree of sensitivity to stress and vulnerability to stress-related illnesses varies between individuals of a given population or even within the same family (Caspi and Moffitt 2006; Rutter 2009, 2010). That may be explained by inheritance of particular genetic factors (genetic variants commonly found in the population) acting as modulators of the outcome of environmental influences. While the idea of a direct causal link between one given gene and one psychiatric disease has been discarded, it is becoming increasingly clear that a combination of several genetic factors can play a significant role in modulating the outcome of environmental influence (Caspi et al. 2010; Caspi and Moffitt 2006; Rutter et al. 2006). For instance, several genetic polymorphisms have been uncovered and found to modulate the sensitivity to stress and vulnerability to stress-related diseases. One of the most investigated gene variants in psychiatry is the functional polymorphism in the promoter region of the serotonin transporter gene (5-HTT; Caspi et al. 2010; Caspi and Moffitt 2006; Rutter et al. 2006). Inheritance of the “short” allele variant results in lower expression levels of 5-HTT in a dose-dependent manner and is associated with increased HPA axis activity. Moreover, carriers of the short allele display increased neuronal activity in amygdala fear pathways and decreased gray matter volume in amygdala and frontal cortex. These alterations appear to be correlated with increased susceptibility to depression and suicide in individuals with a specific history of adverse childhood experiences and/or later stressful life events (Caspi et al. 2010; Caspi and Moffitt 2006; Rutter 2010; Rutter et al. 2006). This represents a powerful demonstration of the impact of genetic variation on the outcome of environmental influences. Several other examples of genetic variation modulating the outcome of childhood and adolescent adverse experiences have been reported (for reviews, see Caspi et al. (2010), Caspi and Moffitt (2006), Rutter (2008), and Rutter et al. (2006)). These include for instance the functional polymorphism in the promoter region of the monoamine oxidase A gene, which moderates the effect of child maltreatment (Caspi et al. 2002; Caspi and Moffitt 2006), and the catechol-O-methyltransferase, which moderates the impact of adolescent cannabis use on development of adult psychosis (Caspi and Moffitt 2006; Caspi et al. 2005).

Importantly, twin and adoption studies have revealed that genetic factors mostly appear to exert significant effects in the context of adverse environmental conditions (Heim and Nemeroff 1999). This is in agreement with the “diathesis–stress model” (Gutman and Nemeroff 2003; Zubin and Spring 1977), which posits that genetic factors that increase vulnerability to stress/diseases are more likely to be expressed under adverse/stressful environments than under more favorable conditions (Barr et al. 2004; Caspi et al. 2010; Caspi and Moffitt 2006; Ouellet-Morin et al. 2009; Rutter et al. 2006). This is well illustrated by the case of the 5-HTT variant described above (Caspi et al. 2010; Caspi and Moffitt 2006; Rutter 2008; Rutter et al. 2006) and in a study of Ouellet-Morin and colleagues (Ouellet-Morin et al. 2009). The authors showed that genetic factors accounted for most of the variation in cortisol levels in infants, as measured by (basal) morning cortisol secretion, only in the context of adverse familial experience (Ouellet-Morin et al. 2009), thus demonstrating the modulating impact of genetic factors on HPA axis development in humans. However, the same group showed that for other measures of stress sensitivity such as HPA reactivity (cortisol) to stress, adverse environmental influences strongly mediated the outcome of HPA axis development with little effects of genetic factors. This suggests that environmental influences can sometimes constrain the modulating impact of genetic factors (Ouellet-Morin et al. 2008). These findings are important for the implementation of intervention programs, where the focus is on attempting to prevent, reverse, and/or reduce familial environmental adversity (Anderson et al. 2003; Doyle et al. 2009).

Furthermore, the spectrum of genetic factors modulating environmental influences not only is restricted to genes coding for proteins but also includes the nonprotein-coding genetic factors (often referred to as junk DNA; Plomin and Davis 2009). Such DNA elements can directly or indirectly influence gene × environment interplays and lead to increased vulnerability to psychiatric disorders (Plomin and Davis 2009; Rutter 2008; Rutter et al. 2006). In addition, DNA copy number variation, where multiple copies of certain genes or segments of DNA are inserted or deleted from particular locations on a given chromosome, represents an untapped source of genetic modulation of environmental influences (Merikangas et al. 2009). Various forms of DNA structural variations have been recently found to play a role in the etiology of psychopathologies such as depression and schizophrenia (Duan et al. 2010; Merikangas et al. 2009; Saus et al. 2010; Weiss 2009; Weiss et al. 2009). It is therefore important to consider all forms of genetic factors as possible modulators of environmental influences on the development of individual differences in stress responsiveness and susceptibility to stress-related diseases.

Genetic modulation in maternal care studies

Also, in animal models, there is clear evidence for genetic modulation of environmental manipulations. For instance, it has been reported that certain strains of mice appear to benefit more from neonatal manipulations such as EH compared to others. This effect can be partially explained by strain-specific differences in maternal mediation (Anisman et al. 1998). Besides substantial variation between dams from the same strain (as previously described in this review), it has been clearly shown that rodents also show strain-dependent variation in maternal care, leading to low and high caring strains (Carlier et al. 1982).

By cross-fostering inbred mice strains to such a low or high caring mother strain, it is possible to determine the contribution of genetic as well as early environmental factors in the development of adult phenotypes (van der Veen et al. 2008). van der Veen and colleagues have investigated this issue. They selected two mice strains based on their maternal care style. One strain showed low (AKR), and the other high (C3H) innate levels of maternal care, therefore providing a significantly different early environment to the offspring. Pups from two unrelated strains (providing different genetic background), namely DBA/2J and C57BL/6J mice, were cross-fostered to these low and high caring mother strains. Indeed, it was observed that AKR dams displayed lower levels of maternal care toward the offspring as compared to C3H dams. However, when the cross-fostered strains were tested in adulthood, a difference in their sensitivity to the “programming” effects of early life environment on adult phenotype was observed. For instance, irrespective of being raised by a low or high maternal care strain, the adult phenotype of C57 mice remained unchanged, indicating a relative resistance to early environmental factors. In contrast, DBA mice raised by a low caring mother strain significantly differed from DBA mice raised by a high caring mother strain in terms of cocaine intake and immobility response in the forced swim test, thus indicating sensitivity to early environmental factors. These results demonstrate a strong effect of genetic background in modulating maternal environmental influences during early life.

Nonshared environmental influences

The outcome of gene × environment interplays described above can also be mediated by more subtle factors, previously underappreciated in their ability to shape the outcome of studies in the field of developmental programming. Even though these factors will add another level of complexity to an already multifaceted field, they should not be neglected. Here, we will focus on the role of nonshared environmental factors on the development of the HPA axis using findings from a recent study in our laboratory.

Shared and nonshared environmental factors

We acknowledge that there are inconsistencies in the way the terms shared and nonshared are used in human as opposed to animal research. The term “nonshared environment” is always used to describe differences in environment. However, in human studies, it is frequently used to describe differences in environment of siblings within a family, for instance due to differential parenting (Feinberg et al. 2000; Reiss et al. 1995). On the contrary, in animal studies, it is used to describe differences between litters such as being raised by a high or low LG dam (Liu et al. 1997). In our manuscript, whenever we mention “shared” or “nonshared”, we meant to describe factors “shared” or “nonshared” by individual members within a litter.

Shared and nonshared environmental factors in developmental programming

Traditionally, the field of developmental programming is focused on investigating the impact of early experiences that differ between litters or families, but are usually considered to be similar (shared) for individual members within the litter or family. Currently, experimental models where the impact of shared and nonshared environmental factors within the litter can be manipulated and/or studied in the laboratory are scarce.

For instance, as described earlier in this review, Meaney and colleagues have convincingly shown the profound and long-lasting impact of naturally occurring variations in early environment (Liu et al. 1997; Macri et al. 2008). However, due to methodological considerations, the association between the early experiences and the phenotype of the offspring in later life is based solely on maternal behaviors (i.e., high vs. low LG dam). Even though the difference between high and low LG offspring is usually referred to as being caused by a nonshared factor, it is assumed that this factor is experienced in a similar way (shared) by all the individual pups within a litter.

However, previous studies by Menard and colleagues using this model show that pups from the same litter display substantial variation in behavioral phenotype later in life (Menard and Hakvoort 2007). This is interesting since it was assumed that these pups were reared and have therefore perceived (shared) the same maternal care environment. This suggests that there are also nonshared factors within the litter that lead to phenotypic variation in the offspring.

In human literature, over the years, an increasing number of studies have indeed reported that parenting style varies not only between families but also within the family. This variation in parenting within the family leads to experiences that are unique to the individual and are nonshared by other members of the family. These nonshared factors are an important correlate for the development of stress-related pathologies in later life (Feinberg et al. 2000; Reiss et al. 1995). Additionally, the same researchers have shown that parenting toward one child is linked with an opposite outcome on the child's sibling as on the target child, suggesting that harsh parenting toward one child could make the other sibling less prone to develop psychopathology. This is referred to as the “sibling barricade” and suggests that the perception of parenting might be more powerful than the absolute amount/quality of parenting you are subjected to (Feinberg et al. 2000; Reiss et al. 1995).

Other human studies by Tremblay and colleagues have shown that in conditions characterized by low familial adversity, differences in cortisol reactivity in young children were mostly accounted for by genetic and nonshared environmental factors. Shared environmental factors did not account for variance in cortisol reactivity (Ouellet-Morin et al. 2008). These findings illustrate the importance of studying the impact of these previously unrecognized subtle nonshared, individual-specific early life experiences within a family or litter, on the development of vulnerability to stress-related disorders.

Modeling shared and nonshared environmental factors in animals

Since experimental models where the impact of nonshared environmental factors within the litter can be manipulated and/or studied in the laboratory are scarce, we have attempted to tackle this issue by investigating an extension of the maternal care model.

To address this issue, we have performed a detailed assessment of the amount and quality of maternal care received by each individual pup within the litter (representing the nonshared influence) during the first week of life. In addition, we also describe the maternal phenotype (representing shared influence). Moreover, we have investigated whether within-litter variation in maternal care can predict phenotypic differences in endocrine responsiveness to acute novelty stress in the offspring later in life. This novel approach is meant to complement previous studies on variation in maternal care between litters and might help to explain gene × environment interplays on an individual level. The methodological details of this experiment can be found in Online Resource 1.

Nonshared environment within the litter predicts stress phenotype in later life

We report that maternal LG is not homogeneously distributed among individual pups within the litter, suggesting that particular pups consistently receive higher or lower levels of LG compared to their littermates (Fig. 1).

Fig. 1
figure 1

Distribution of maternal LG received by individual pups for each litter (1–7). Shown is the percentage of observations (pnd 1–7) in which each individual pup received maternal LG. Male pups are indicated with male signs, and female pups are indicated with female signs. “Highs” and “lows” within the litter (that displayed a LG percentage of at least one standard deviation above or below the family mean, respectively) are indicated with enlarged male or female signs. Inset Total percentage of LG displayed by each mother

Full size image

Next, we examined whether individuals that received low and high levels of maternal LG acquired through nonshared experience differ in their neuroendocrine response to acute stress. For this purpose, animals were exposed in adolescence (pnd 28) and adulthood to an acute novelty stressor (10-min exposure to a novel open field). We report that, both in adolescence (Fig. 2a) and adulthood (Fig. 2b), glucocorticoid levels were significantly more elevated in low LG offspring when compared to high LG offspring in response to acute stress. These findings suggest that enduring changes in endocrine response to novelty stress can be “programmed” via variation not only in shared but also in nonshared individual early life history (i.e., individual level of maternal LG). For a detailed description of the methodology and supplementary data of the experiment on stress responsiveness, see Online Resource 2.

Fig. 2
figure 2

Plasma corticosterone (CORT) levels of rats characterized as high or low LG within the litter, expressed as mean ± SEM. *p < 0.05; **p < 0.01. Selected high and low LG animals (n = 13) were tested for their endocrine responsiveness to acute novelty stress in (a) adolescence (age pnd 28) and (b) adulthood (age 4.5 months). CORT levels were significantly more elevated in low LG offspring when compared to high LG offspring in response to acute stress. Due to technical issues, we are unable to show basal CORT levels of the animals at adulthood; therefore, in b, only stress-induced CORT levels are displayed

Full size image

Additional considerations

These findings are reminiscent of previous maternal care studies describing the impact of shared maternal LG on the endocrine response to acute stress (Liu et al. 1997). However, the current findings suggest that very subtle differences in nonshared maternal LG (that are up to tenfold smaller than previously reported differences in LG used to characterize dams within a cohort (Champagne et al. 2003)) appear to have comparable predictive value for later life stress responsiveness. Definitely, further studies are warranted to understand these observations. However, the following factors should be taken into consideration when designing an experiment on this topic.

Features of maternal care repertoire

Although all features of the maternal care repertoire can be measured using this paradigm, it is not always possible to accurately measure their distribution over the individual pups. Arched-back nursing (ABN) is a clear example of this limitation. ABN is often scored in co-occurrence with maternal LG, but can also be independently affected by early experiences such as EH and can as a consequence exert differential effects on the offspring (Pryce et al. 2001). However, since pups are usually nursed as a group and differences in nursing posture (low–high kyphosis) affect the whole litter, it is unlikely that variation in ABN underlies within-litter differences in later life phenotype. In our opinion, ABN can be better considered as a shared environmental influence for the pup (Champagne et al. 2003) that is not likely to account for the findings reported here. Nonetheless, we acknowledge that shared maternal factors (i.e., ABN style of dam) play a role in the development of individual phenotype, although their effects might be modulated by nonshared factors (i.e., individual differences in maternal LG).

Introduction of early handling

Secondly, in order to score individual LG (nonshared factor), a daily marking procedure was used to label and identify individual littermates. Such a procedure inevitably introduces a substantial amount of EH. This implies that besides studying the impact of subtle differences in maternal care within the litter, one has to take into account the well-known effects of EH (see Environmental influences section). One possible bias is that our daily EH/marking procedure potentially altered maternal care directed toward the offspring. However, we believe that this bias is distributed equally and to a similar extent in all litters studied, and is not likely to account for the effects observed here. Furthermore, since we observed a substantial amount of variation between littermates in terms of stress responsiveness, we suggest that nonshared differences in maternal care rather than EH/marking mediate the effects observed here. This scenario is not unprecedented and has been previously proposed by Macri and coworkers for variation in shared maternal care (Macri et al. 2008; Macri et al. 2004).

Underestimation of individual differences in LG

The Wistar dams displayed relatively low levels of maternal care when compared to previously described maternal LG distribution range in the Long–Evans strain (Champagne et al. 2003). Therefore, the average levels of LG (i.e., 6%) displayed by Wistar dams in our study fall in the range that would be equivalent to levels displayed by low LG Long–Evans dams. In our study, we observed that on average, Wistar dams delivered 40 LG events over the 7-day observation period. From the pup perspective, this might represent on average five LG events per individual pup. Such score might appear too low to reliably categorize individual littermates as low or high LG offspring. We acknowledge that this represents a limitation of our experimental approach, which leads to an underestimation of the occurrence of pup-directed LG events. In the future, this issue can be resolved by increasing the amount of observations per day to achieve a more reliable representation of the actual differences in nonshared maternal LG.

Gender bias

It is noteworthy that for most of the litters studied, dams displayed a strong gender preference for male over female offspring. Such a preference of the mother for male pups has been shown before (Deviterne and Desor 1990; Moore and Morelli 1979) and proposed to be regulated by testosterone, pheromones, and urinary odor (Moore 1985, 1986). As a consequence, this gender preference resulted in a bias in the distribution of males and females over the selected low (exclusively females) and high LG (almost exclusively males) groups. Since glucocorticoid levels and corticosteroid-binding globulin levels in the blood are modulated by fluctuations in sex hormone levels (McCormick et al. 1998), one has to be careful when interpreting the association between early life maternal LG and later life HPA axis activity. However, we report an association between neonatal LG and glucocorticoid levels as early as adolescence (pnd 28), when the regulatory role of sex hormones on HPA axis functioning is assumed to be low compared to that in adulthood (McCormick and Mathews 2007). Gender differences in endocrine responsiveness at pnd 28 have been shown in some (Hary et al. 1981; Hary et al. 1986) but not all (Sencar-Cupovic and Milkovic 1976) studies. Because of the small sample size, we were unable to reveal a within-gender effect of maternal (nonshared) LG on glucocorticoid levels. Future studies using a higher number of animals and including ovariectomized female offspring will reveal the exact role of nonshared gender-dependent differences in maternal LG during early life on the development of stress phenotype in later life.

For supplementary material on the gender-dependent differences in maternal care, see Online Resource 2.

Conclusion

In this review, we have provided an overview of some of the scenarios in which environment and genetic variation play a role in the development of individual differences in sensitivity to stress and vulnerability to diseases. Four major points emerge with significant relevance for the field of developmental programming: (1) although environmental influences are recognized as major contributors to the development of stress sensitivity and vulnerability to psychopathology, they do not always lead to the development of diseases; (2) considerable heterogeneity exists between individuals in their response to these environmental factors; (3) complex gene × environment interplays are hypothesized to account for this phenomenon, a theory gaining momentum in psychiatry; and (4) understanding of the mechanisms by which gene × environment interplays underlie stress sensitivity and disease vulnerability requires the study of their biological consequences as well as identification of the relevant biological pathways (Caspi et al. 2010; Caspi and Moffitt 2006; Rutter 2008, 2010; Rutter et al. 2006).

The several forms of genetic variation do little on their own to cause or alter basic biology and pathways to diseases in response to environmental influences (Rutter 2008, 2010; Rutter et al. 2006). Genetic variation can only exert significant effects via its impact on gene and protein expression, which determines the biologically relevant outcome. Inheritance of particular genetic variants represents one way by which gene and protein levels can be modified. However, environmental influence can also modulate gene expression via epigenetic mechanisms (Weaver et al. 2004). While the fields of developmental programming, neuroscience, and psychiatry have made significant breakthrough on all fronts by uncovering the several ways by which environmental risk factors and genetic variants operate through biological pathways, these fields are still confronted with significant methodological challenges. One aspect of the methodological challenge lies in the ability to develop animal models to address the impact of nonshared environmental influences on the outcome of gene × environment interplays. The original findings presented in this article suggest that nonshared influences on the development of individual differences in sensitivity to stress can be studied in the laboratory.

Furthermore, from an evolutionary perspective, gene × environment interplays and their related biological mechanisms leading to “programming of the HPA axis” (i.e., alterations in stress sensitivity) are generally meant to be adaptive and not necessarily a substrate for diseases. This is the basis of the “predictive adaptation plasticity hypothesis” (Crespi and Denver 2005; Gluckman and Hanson 2007; Gluckman et al. 2007; Gluckman et al. 2010; Horton 2005). This theory is based on the concept that a developing organism responds to cues (e.g., maternal LG) in its environment by changing certain aspects of its homeostasis (e.g., HPA axis) in order to produce a phenotype that is highly adapted to its current and future environment. This concept led to a currently dominant view in medicine stating that a high degree of “mismatch” between the early and later life environments confers an increased risk to develop adult diseases (Gluckman and Hanson 2007; Gluckman et al. 2007; Gluckman et al. 2010; Horton 2005). There is much evidence to support this view in the field of metabolic and cardiovascular disorders (Gluckman et al. 2007; Prentice 2005). However, in psychiatric research, the validity of this concept is uncertain.

Recent evidence from animal studies suggests that the concept of “mismatch” can also apply to the development of individual differences in stress sensitivity. We have recently shown that the outcome of early experience on stress-related parameters is dependent on later life context (Bagot et al. 2009; Champagne et al. 2008). Specifically, we reported that adult offspring of low LG mothers (considered as a form of adversity) show indeed the “expected” low cognitive performance in a low-stress context. However, in a high-stress context, their performance was better when compared to animals that had received high levels of maternal care, which in turn were impaired under the same stressful conditions (Bagot et al. 2009; Champagne et al. 2008). These findings suggest that the influence of shared and possibly nonshared environmental experiences (as presented in this review) during development might serve as a basis for resilience to stressful challenges in later life in the context of “matched” environments. However, whether such phenomenon plays a role in the modulation of gene × environment interplays leading to increased susceptibility or resilience to stress-related diseases remains to be further ascertained.