SEARCH

SEARCH BY CITATION

Keywords:

  • birth weight;
  • prenatal stress;
  • fetal growth;
  • anxiety;
  • dexamethasone

Abstract

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References

A large body of human epidemiological data, as well as experimental studies, suggest that environmental factors operating early in life potently affect developing systems, permanently altering structure and function throughout life. This process with its persistent organizational effects has been called ‘programming’. The brain is a key target for such effects. This review focuses on the effects of adverse early environments, notably exposure to stress or glucocorticoids, upon subsequent adult hypothalamus-pituitary-adrenal axis activity, behaviour and cognition. We discuss the effects observed, the proposed underlying molecular and cellular mechanisms and the consequences for pathophysiology. The data suggest that key targets for programming include glucocorticoid receptor gene expression and the corticotrophin-releasing hormone system. Increasing evidence for analogous processes in humans is also reviewed. Early life programming of neuroendocrine systems and behaviour by stress and exogenous or endogenous glucocorticoids appears to be a fundamental process underpinning common disorders. Approaches to minimize or reverse the consequences of such early life events may have therapeutic importance.

A host of recent epidemiological studies in diverse human populations have suggested that factors acting during early life are associated with the development many decades later of common adult cardiovascular and metabolic disorders. These studies have shown that low birth weight or thinness at birth predicts a substantially increased risk of adult hypertension, impaired glucose tolerance/type 2 diabetes mellitus, hyperlipidaemia and deaths from coronary heart disease (1–4). Classical risk factors in adulthood (smoking, obesity, diet, alcohol, physical inactivity) appear additive to the early life influences, suggesting the early life effect has distinct causes and roles. The link between prenatal growth retardation and adult cardiovascular and metabolic disorders has also been found in specific animal models of restrained fetal growth (5–9). These intriguing findings have spawned the ‘fetal origins’ hypothesis of adult disease (10–12).

It is unlikely that low birth weight per se causes these increased risks of adult disease. Rather, there may be a common factor that influences both intrauterine growth as well as the set-point of adult physiological systems. Some data suggest that both fetal growth and later pathophysiology may be influenced by the same gene. Examples include genes involved in insulin homeostasis, such as the ‘glucose sensor enzyme’ glucokinase, which might affect fetal growth and also cause the adult ‘metabolic syndrome’ phenotype of insulin resistance (13). However, most data suggest that the ‘fetal origins’ observations, at least in part, reflect the actions of environmental factors. The concept that nongenetic factors act early in life permanently to organize or imprint physiological systems is known as perinatal ‘programming’. This concept is not new, but has been a long-standing topic in neuroendocrinology research. For example, sexually dimorphic structures and functions in the vertebrate hypothalamus are programmed by sex steroids, and this has been the subject of intensive investigation for many decades [for a review, see (14)]. It is revealing that such perinatal sex steroid exposure also permanently programmes enzymes in the liver (15).

The biological ‘purpose’ of early life programming is not known. However, it can be speculated that prenatal plasticity of physiological systems allows environmental factors, acting on the mother and/or the fetus, to alter the set-point or ‘hard-wire’ the differentiated functions of an organ or tissue system to prepare the unborn animal optimally for the environmental conditions ex utero. If the environmental circumstances in later life are not as anticipated, such perinatal programming might produce maladaptive physiology and ultimately predispose to disease. Research into the link between fetal growth and adult disease has largely focused on metabolic and cardiovascular disorders. But is the central nervous system (CNS) also involved?

Programming and the CNS

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References

The brain is very sensitive to prenatal programming. A number of agents are known permanently to affect brain development, such as growth factors, transcription factors and nutrients. Steroids in particular have powerful brain-programming properties (14). One of the most intensively studied programmed CNS systems is the hypothalamic-pituitary-adrenal (HPA) axis. The axis mediates the release of glucocorticoids to diurnal cues and stress and is outlined in Fig. 1. In humans, the main glucocorticoid is cortisol, whereas in rats and mice it is corticosterone (CORT). Glucocorticoids act predominantly via intracellular receptors, which function as ligand-activated transcription factors. There are two receptor subtypes: the lower affinity glucocorticoid (GR) and higher affinity mineralocorticoid (MR) receptor. Glucocorticoids regulate their own secretion by negative feedback actions at the level of the hypothalamus and the pituitary, inhibiting the synthesis and/or release of corticotropin releasing hormone (CRH), arginine vasopressin (AVP), proopiomelanocortin (POMC) and corticotropin (ACTH), thus terminating the stress response (Fig. 1). Further control of HPA activity takes place at extrahypothalamic sites, notably the hippocampus (16) and the amygdala (17, 18). The HPA axis is particularly susceptible to early life programming. Early postnatal events that disturb mother–pup interactions permanently alter the activity of this system. For example, in the first few days of life short-term (15 min) daily separation of pups from the mother, also called neonatal handling (19), as well as longer-term (3–24 h) maternal deprivation (20, 21) produce specific and permanent changes in offspring plasma ACTH and CORT concentrations (19–21) and corticosteroid receptor levels in hippocampus (19, 22), PVN and pituitary (21). But what about prenatal events? A paradigm frequently used to investigate early life manipulations of the HPA axis is that of prenatal stress. Here, we review the evidence that prenatal stress programmes brain function, notably of the HPA axis, and then address the role of fetal glucocorticoid exposure in this important biological effect.

image

Figure 1. Schematic representation of the hypothalamic-pituitary-adrenal (HPA) axis. GR, glucocorticoid receptor; MR, mineralocorticoid receptor; PVN, paraventricular nucleus; CRH, corticotropin-releasing hormone; AVP, arginine vasopressin; ACTH, adrenocorticotropic hormone.

Download figure to PowerPoint

Prenatal stress

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References

In this paradigm, pregnant dams are exposed to stress, typically one to three times daily, either throughout or during a specific period of pregnancy. This evokes a rise in plasma ACTH and CORT levels in the dams (23, 24), although in the last third of pregnancy the response may be reduced compared to that in virgin rats (25). Broadly, such studies have shown that maternal stress has permanent and often profound effects on the offspring. Brain systems targeted by prenatal stress include hypothalamic structures involved in the control of sexual behaviour (26), causing demasculinization of sexual behaviour in male offspring (27). Furthermore, prenatally stressed rats display altered, usually exaggerated, neuroendocrine and behavioural responses to stress (see below) (28–31). There is, however, considerable variation in the magnitude of the persisting effects upon the offspring and some data are even contradictory. For example, in some studies, prenatal stress only affects HPA activity in females (32), whereas in others changes are seen in males (31). This is probably due to substantial variability between studies in the type of stressor used, its gestational timing and the duration and predictability of the stress event(s) (33–35). Furthermore, strain differences in the programming effects of prenatal stress (36), suggest that potent genetic constraints operate too.

Effects on HPA activity

Prenatal stress has long-lasting effects on the adult offspring HPA-axis, in general programming a persistently hyperactive system. To address the possible pathogenesis of such effects, several studies have examined the postnatal ontogeny of changes in HPA activity following prenatal stress. Investigation of the effects of prenatal stress upon basal HPA-activity during the first weeks of life has given variable results, depending on the type and timing of the maternal stress. In some cases, male prenatally stressed rats have elevated basal CORT levels (35, 37) up until weaning, whereas others found no effect of prenatal stress on basal CORT (31, 38, 39) or ACTH (40) during this time. In contrast, stress-induced rises in CORT and ACTH are consistently higher in prenatally stressed animals throughout the preweaning period (37–39), as are poststress CORT levels (28, 31, 37). Such changes in HPA reactivity interfere with the stress hyporesponsive period normally seen in rats between postnatal day (pnd) 4 and pnd 14 (41). Hippocampal MR and GR binding is lower in prenatally stressed rats at the time of weaning, but not on pnd 3 (31). Increased plasma CORT levels (which are already observed on pnd 3) precede such changes in hippocampal MR and GR binding. In prenatally stressed rats, it is possible that the apparent absence of a stress hyporesponsive period, and thus elevated HPA activity in neonatal life, may underlie the altered development of hippocampal MR and GR density, although (negative) autoregulation of GR does not appear to show full function until after weaning (42, 43). Other possible mechanisms are discussed below.

In adult rats, the normal rise in plasma CORT levels induced by stress is increased in male and female rats (31, 32) exposed prenatally to stress. This effect is most marked in the later poststress samples, suggesting impaired negative-feedback regulation of the HPA axis (31, 32, 44). Such HPA programming effects of prenatal stress effects are usually more marked in female than male offspring (32, 45–47). Indeed, basal (unstressed, diurnal nadir) plasma CORT and ACTH levels in adult, prenatally stressed rats are usually reported as unchanged in males (29, 30, 32, 36, 45, 48, 49), whereas these basal parameters are increased in females (32, 45).

Several groups have shown that prenatal stress causes alterations in adult hippocampal corticosteroid receptor density. Again, there are often sex differences, with permanently decreased numbers of hippocampal MR in prenatally stressed males (30, 31, 44, 48), whereas in females the lower affinity GR is decreased (32). Hippocampal MR are involved in the control of basal HPA activity and, with GR, coordinate negative-feedback regulation after stress (50, 51), suggesting that permanently reduced hippocampal MR density may be responsible for the increased poststress CORT levels observed in adult rats exposed prenatally to stress.

Effects on behaviour

Anxiety

Prenatally stressed adult rats have often been described as having a higher degree of ‘emotionality’ (38). When exposed to an ‘open field’, prenatally stressed rats typically show decreased locomotion and increased defecation (38, 52–54). However, higher initial open field activity has also been observed (30). These atypical results were explained as an increased level of ‘escape behaviour’, coincidentally underlining the difficulty in interpreting animal behaviour. Avoidance of anxiogenic locations in adult prenatally stressed animals has also been shown, indicated by a low number of visits to the open arms on an ‘elevated plus’ maze (30, 52) which in one study was correlated with plasma CORT levels (30). The prenatally stressed rat also exhibits a reduced propensity to play (55), an increase in defensive freezing (29), reduced movement in an activity wheel (56) and increased ultrasonic vocalizations in an open field (57). Overall, these data show that prenatal stress programmes adult behavioural suppression or increased anxiety in aversive situations. The amygdala mediates anxiety and fear-related behaviour and learning (58, 59), and contains CRH-positive cells as well as CRF receptors (59). Indeed, CRH injection into this structure promotes behaviours reminiscent of anxiety (58), suggesting that CRH may be the key neurotransmitter in these effects. It is therefore tempting to speculate that the increased amygdala content of CRH observed in prenatally stressed rats (60) may underlie their apparent ‘hyper-emotional’ state and reduced coping ability. Interestingly, depression in humans is associated with elevated CRH levels in cerebrospinal fluid (61), supporting the notion that this neuropeptide plays a key role in anxiety and affective disorders (62).

Cognition

Adult cognitive abilities are clearly affected by prenatal stress. Both acute and repeated maternal stress reduces learning by adult offspring in an operant discrimination task (63). Furthermore, prenatal stress alters offspring behaviour in the watermaze. This is perhaps not surprising, since watermaze performance is a hippocampus-associated function (64) which is affected by corticosteroid receptor occupation (65–67), and prenatally stressed animals have reduced levels of hippocampal corticosteroid receptors (31, 32). Thus, although they do not differ from control rats in the learning phase of the task, prenatally stressed rats spend more time looking for the platform in a reversal task in a watermaze (68, 69), indicating a different, perhaps more rigid search strategy. Although a recent study failed to find any effects of prenatal stress on watermaze performance (30), the test conditions may be crucial to discriminate effects. In particular, CORT secretion during the learning phase of this test may influence retention (memory) (70), since only training in stressful conditions (cold water, 12 °C) revealed differences between prenatally stressed and control rats (68). Other tests of cognitive functioning support the general watermaze findings (71), but overall there are relatively few studies that directly describe the effects of prenatal stress on cognitive function. Considering the increased HPA activity in prenatally stressed rats, the ‘glucocorticoid cascade hypothesis’ described by Sapolsky et al. (72) would predict that hippocampal damage as a consequence of cumulative exposure to elevated CORT levels occurs earlier in prenatally stressed rats than in controls, and this would show in cognitive tests associated with the hippocampus. Indeed, old rats exposed prenatally to stress show worse working memory in the radial maze and exhibit memory deficits much earlier in midlife (15 months) than control rats when tested in a Y-maze (71). Thus differences between prenatally stressed and control-reared animals may become more apparent with ageing.

Neural mechanisms

Because brain neurotransmitter systems and glucocorticoids interact to modulate both behaviour and HPA activity (73), it is possible that the effects of prenatal stress are mediated by permanent alterations in these systems. Indeed, prenatal stress reduces 5-HT, noradrenaline and dopamine levels and turnover in the adult brain (29, 39, 69). Prenatal stress also permanently increases hippocampal acetylcholine release in response to stress or CRH injection (74). Furthermore, adult prenatally stressed rats show decreased hippocampal synaptic density, a parameter modulated by 5-HT and other neurotransmitters (69). While the systems underlying the effects of prenatal stress are poorly understood, more light has been shed on postnatal environmental programming of the hippocampal GR/MR system. Disturbing mother-pup relationship by short daily separations (‘neonatal handling’) during the first 3 weeks of life permanently increases GR levels in the hippocampus of pups (21). The mechanism involved has, in part, been elucidated, acting through altered 5-HT neurotransmission, mediated in part via increases in circulating thyroid hormones (75–77). In this paradigm, 5-HT, acting via specific 5-HT receptors on hippocampal neurones, induces a second messenger cascade involving cAMP and protein kinase A. Thereafter there are changes in cAMP-associated and other transcription factors/third messengers, notably AP-2, NGFI-A, which might activate specific GR gene promoters expressed in the hippocampus (78, 79). Whether a similar mechanism modulates the effects of other perinatal manipulations is not known, although it is interesting that the reduced growth seen in maternally deprived rats results from a suppression of growth hormone secretion mediated by 5-HT2A receptors (80).

Summarizing, there is substantial evidence to suggest that prenatal stress programmes the HPA axis as well as behaviour, and that plasticity of developing brain monoamine systems underlies, in part, these changes. Ultimately, prenatal stress results in adults that exhibit impaired coping and increased anxiety. Intriguingly, in humans, prenatal stress has been suggested to increase the risk of schizophrenia (81), a putative neurodevelopmental disorder itself associated with alterations in brain monoamine levels. Moreover, disturbances in HPA regulation and brain monoamine levels have been associated with affective and anxiety disorders in humans (61, 82–84), suggesting that such conditions may be (partially) prenatally programmed. Indeed, several psychological and behavioural abnormalities have been reported in children exposed to ‘prenatal stress’ (85–87). The mechanisms underlying the effects of prenatal stress have not yet been convincingly established. In the search for a ‘programming factor’ imprinting on the development of fetal tissues, two major hypotheses have arisen which will be discussed below.

Mediators of early life programming

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References

Glucocorticoids

Because an important feature of the stress response is the secretion of high levels of CORT, this steroid has become an obvious candidate for the role of ‘programming factor’ in the prenatal stress paradigm. Moreover, prenatal glucocorticoid exposure has recently been implicated in the development of adult hyperglycaemia and hypertension (9, 88), and may represent the epidemiological link observed between these disorders and low birth weight (1, 3). Experiments describing the effects of injecting pregnant rats with dexamethasone (DEX), CORT or ACTH have shed some light on the involvement of prenatal glucocorticoid exposure in fetal programming of the brain.

Daily administration of ACTH in the last week of pregnancy alters adult brain monoaminergic activity and reduces adrenal weight in combination with elevated basal CORT, but reduces stress-induced CORT levels in plasma (46). As with prenatal stress, prenatal ACTH injections have more pronounced effects in female than in male rats (46). Because ACTH does not pass the placental barrier (89), prenatal ACTH has been assumed to act by increasing maternal CORT levels. Several studies have aimed directly to assess the effects of prenatal exposure to high CORT concentrations. In one study (90) rats were given CORT-pellets (releasing 28 mg/day) in the second week of pregnancy. At 21 days of age, the offspring had unaltered basal plasma CORT and ACTH levels, but decreased stress-induced CORT levels (68). A second study (27) reported unaltered stress-induced plasma CORT levels in 90-day-old offspring from rats that were injected three times daily with 7 mg/kg CORT in the third week of pregnancy (gestation is approximately 3 weeks in the rat). In contrast, a third study administered a much lower dose of CORT (3 mg/kg) to adrenalectomised dams before daily restraint procedures and observed effects in the offspring identical to those of prenatal stress, with a prolonged stress-induced CORT response and decreased hippocampal MR numbers (48). In this study, maternal stress had no effects in offspring of adrenalectomized pregnant rats that were replaced with constant low-dose CORT levels. Although the authors concluded that stress-induced increases in maternal glucocorticoids may be the mechanism by which prenatal stress impairs the adult offspring HPA response, it appears more likely that the combination of stress and high CORT levels is critical for the observed effects. Thus, an interaction between CORT and other stress-induced factors (e.g. neurotransmitters, opioids, neurosteroids, circulating hormones) may be required to induce long-term changes in the HPA axis of prenatally stressed animals. Indeed, blocking of opioid receptors at the time of the prenatal stress decreased the anxiety of prenatally stressed rats in the elevated plus-maze later in life (91).

Studies with the GR agonist dexamethasone

A large number of animal studies have described the effects of prenatal exposure to the synthetic glucocorticoid DEX, which relatively readily passes the placenta, and is used in clinical obstetric practice. At conventional therapeutic doses, DEX is a potent GR, but not MR agonist. Decreased brain weight is a well-established consequence of perinatal DEX administered in doses varying from 0.1 to 3 mg/kg (27, 92–95). Prenatal DEX exposure in late gestation (embryonic days 17, 18 and 19) influences brain development at a dose as low as 0.05 mg/kg, altering the induction of nuclear transcription factors such as c-fos and AP-1 (96). Furthermore, it consistently reduces birth weight (6, 27, 88, 95, 97, 98), reflecting not only the reduction in brain weight but also (mostly transiently) reduced body weight (92, 93, 95, 98).

Prenatal DEX exposure also alters adult behaviour. In recent studies we found that administration of 0.1 mg/kg DEX either throughout the three weeks of gestation (DEX1-3) or in the last week of pregnancy only (DEX3) reduces ambulation and rearing in an open field in adult rats (Fig. 2). However, another study showed that exposure during the third week of gestation to the same dose of DEX did not alter open field behaviour (27), whereas administration of 0.05 mg/kg on gestational days 17, 18 and 19 increased activity and reduced defecation (98). These data suggest perhaps that very specific time windows exist for the effects of prenatal treatments, as seen previously in prenatal stress procedures. Indeed, only late gestational DEX (0.1 mg/kg) alters exploration on an elevated plus-maze (Fig. 2). Furthermore, only DEX3 offspring have reduced immobility both in the acquisition and the retrieval phase of a forced-swim test, indicating impaired coping and a reduced capacity for acquisition, consolidation and/or retrieval of information under stressful circumstances (Fig. 2). This suggests that fetal glucocorticoid exposure, especially in the last week of gestation, may programme ‘behavioural inhibition’ and reduced coping in aversive situations later in life. Prenatal DEX (0.05 mg/kg) also alters male sexual behaviour, as does prenatal stress (69). Prenatal CORT or ACTH do not replicate these behavioural effects (27), suggesting DEX may affect parts of the developing fetal CNS that CORT cannot access (see below).

image

Figure 2. Effects of prenatal exposure to dexamethasone (DEX) throughout gestation (DEX1-3) or in the last week of gestation only (DEX3) on behaviour. Open bars, open-arm entries on elevated-plus-maze; Striped bars, time spent floating in a forced-swim test; Closed bars, number of rears in a 12-min open-field test. *P < 0.05, **P < 0.01.

Download figure to PowerPoint

The behavioural changes observed in DEX offspring may be associated with altered functioning of the amygdala. Prenatal DEX (0.1 mg/kg) treatment in both continuous and last trimester paradigms increases CRH mRNA levels specifically in the central nucleus of the amygdala (CEA), a key locus for the effects of the neuropeptide upon the expression of fear and anxiety. Several studies have shown that corticosteroids facilitate CRH mRNA expression in CEA (99, 100). Indeed, prenatal DEX also increases GR and MR in the amygdala, albeit in the basolateral nucleus (BLA), which projects to the CEA, rather than in the CEA itself. Since CORT levels are maintained or even elevated in the adult offspring of DEX-exposed pregnancies (6, 27), these findings suggest that an elevated corticosteroid signal in the amygdala may produce the increased CRH levels seen in adulthood (Fig. 3). Direct tests of this assertion are awaited.

image

Figure 3. Schematic representation of changes taking place in prenatal dexamethasone (DEX)-exposed rats. The glucocorticoid signal in the amygdala is elevated as a result of increased plasma CORT (corticosterone) levels and increased expression of corticosteroid receptors in the BLA (basolateral nucleus of the amygdala). The BLA projects to the CEA (central nucleus of the amygdala), where an increased glucocorticoid signal unpregulates the expression of CRH (corticotropin-releasing hormone) in the CEA, resulting in increased anxiety and reduced coping. GR, glucocorticoid receptor; MR, mineralocorticoid receptor.

Download figure to PowerPoint

A direct relationship between brain corticosteroid receptor levels and anxiety-like behaviour is supported by the phenotype of transgenic mice with disrupted GR expression in the brain, which show strikingly attenuated anxiety (101). Central or intra-amygdaloid administration of CRH is anxiogenic [for a review, see (102)], whereas absence of the CRH1 receptor reduces anxiety-like behaviour (103), and it may be postulated that high CEA CRH levels in rats exposed to glucocorticoids in utero cause increased ‘allostatic load’[allostasis being stability that is maintained through anticipatory regulation of many variables, see (104)], which is expressed behaviourally in adverse environments.

The programming window?

Prenatal DEX effects on HPA activity also appear to depend on the timing of administration. Both long-term and short-term (6, 98) exposure to 0.1 mg/kg DEX in utero result in adults with elevated basal CORT levels. However, some preliminary data suggest that the mechanism underlying the increased HPA activity may differ depending upon the timing of exposure. DEX exposure in the last third of gestation increases CRH mRNA in the hypothalamic PVN, and reduces MR and GR levels in the hippocampus (6), whereas long-term DEX treatment throughout gestation does not alter hippocampal MR or GR, but increases corticosteroid receptor expression in the amygdala, which exerts predominantly a stimulatory effect upon the HPA axis (17, 18). The implications are that late gestational exposure to DEX may permanently alter the ‘set point’ of the HPA axis at the level of the hippocampus, reducing feedback sensitivity and thus elevating basal HPA activity. In contrast, continuous exposure to DEX through gestation may increase forward drive of the HPA axis, perhaps by increasing amygdala sensitivity to glucocorticoids, although there is currently no direct evidence for such a glucocorticoid-sensitive ‘feed-forward’ system. It is clear that DEX exerts permanent developmental stage-and locus-specific effects upon GR and MR expression in the brain.

Although the mechanisms that underlie the time specificity of DEX effects are unknown, there appear to be several distinct time windows during which glucocorticoid exposure has different effects. This is not an uncommon phenomenon. For example, prenatal (maternal stress) and neonatal (handling) manipulations have opposite effects on hippocampal corticosteroid receptor binding and stress-induced HPA activity (19, 31). Thus, the issue arises of why specific CNS systems are programmed during particular exposure windows and by specific manipulations. Here, we may assume that this depends critically on (i) the local expression of MR and GR, which have distinct developmental patterns of expression (105, 106); (ii) access of the programming agent to the tissue (e.g. expression of the potent CORT-metabolizing enzyme 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2, see below) in the developing CNS exhibits very complex patterns that presumably determine whether CORT – but not DEX – ever gets to the local GR (107, 108); and, perhaps most importantly (iii) on the state of differentiation of that tissue. It seems reasonable to propose that programming can only happen during the development of a tissue. Before initial development or after terminal differentiation the tissue is, almost by definition, refractory to programming effects. Thus glucocorticoid exposure in, for example, the last days of gestation in the rat can target CNS regions actively developing, such as the hippocampus, but not those yet to develop or already in their final state. Of course the long and complex pre- and postnatal ontogeny of the brain make it a prime target for programming. In support of this schema, prenatal DEX exposure throughout gestation has distinct effects on the HPA axis from DEX merely in the last week of pregnancy: the latter predominantly programming GR and MR expression in the hippocampus, the former in the amygdala (6). Interestingly, a critical time window in the last week of gestation has also been observed for prenatal DEX programming of glucose intolerance in rats (9), whereas adult blood pressure is elevated following either DEX throughout pregnancy (88) or only in the last third of gestation (6). In contrast, in sheep, both hypertension and increased insulin sensitivity are programmed by DEX during critical periods proportionately earlier in gestation (109).

Neural mechanisms

The superficial commonalities between the effects of prenatal (stress or DEX exposure) and postnatal (handling or maternal deprivation) environmental manipulations appear to reflect distinct underlying mechanisms. Similar to prenatal stress, prenatal DEX exposure, mostly in the last third of gestation, permanently alters developing monoaminergic systems. The 5-HT system seems particularly sensitive, with DEX decreasing brain 5-HT levels and turnover and altering the density of 5-HT reuptake sites at the synapse (98, 110). As mentioned above, postnatal manipulations such as handling also affect the 5-HT system, increasing 5-HT turnover in the hippocampus (76). This again illustrates how distinct programming mechanisms operating at different times of development can produce similar (but opposite) and permanent alterations in phenotype. Following neonatal handling, events downstream of 5-HT receptors may ultimately permanently affect neuronal gene expression, such as persisting changes in levels of transcription factors, including AP-2 and NGFI-A (78, 79). Whether such changes also play a role in prenatal DEX programming remains to be determined.

Summarizing, prenatal DEX exposure, particularly in late gestation, clearly has programming capacities, acting centrally as well as peripherally (6, 9, 65). The effects of prenatal high endogenous/physiological glucocorticoid levels are less clear-cut and this raises the question of whether maternal glucocorticoids reach the fetus to the same extent as DEX does. This question will be discussed in more detail in the next section.

11β-Hydroxysteroid dehydrogenase, the feto–placental barrier to glucocorticoids

It has been suggested that maternal glucocorticoids at least partly underlie the effects of prenatal stress on the fetus (48). The fact that steroids are highly lipophilic and therefore easily cross the placenta has contributed to this hypothesis. Indeed, many studies into the consequences of prenatal glucocorticoid exposure refer to a 1970 study by Zarrow et al. (111), which appears to show the passage of 14C-CORT from the rat mother to the fetus. However, this study, in which mothers were injected with 14C-CORT, only measured the amount of radioactivity that had passed from mother to fetus, assuming it was still incorporated into CORT. The possibility that part of the radiolabel might have been incorporated into CORT-metabolites was not considered.

It is now well established that a potent, high-affinity CORT-metabolizing enzyme is expressed in the placenta (107, 112, 113), as well as a few other specific tissues (114, 115). 11β-hydroxysteroid dehydrogenase (11β-HSD) catalyses the interconversion of CORT (cortisol in humans) and 11-dehydroCORT (cortisone) (Fig. 4). The latter, 11-keto metabolites, have a very low affinity for corticosteroid receptors and are considered inert. 11β-HSD is thought to regulate corticosteroid access to its receptors. Two distinct isozymes of 11β-HSD have been characterized. The type 1 enzyme (11β-HSD1) is a widespread NADP(H)-dependent enzyme (116, 117) with a relatively low affinity for CORT (Km in low micromolar range). In intact cells, including neurones, 11β-HSD1 catalyses predominantly the 11β-reductase reaction, regenerating active forms (118, 119), or is bidirectional (117). In contrast, the type 2 isozyme (11β-HSD2) is an NAD-dependent, exclusive dehydrogenase which inactivates CORT for which it has a very high affinity (Km in low nanomolar range) (112, 120). It is 11β-HSD2 that is abundantly present in the placenta, notably the trophoblast (121).

image

Figure 4. Interconversion of corticosterone and 11-dehydrocorticosterone by 11β-HSD. The type 1 enzyme (11β-HSD1) reactivates 11-dehydrocorticosterone by catalysing the reductase reaction using NADPH as cofactor, whereas the type 2 enzyme (11β-HSD2) acts as a dehydrogenase and inactivates corticosterone, using NAD as cofactor.

Download figure to PowerPoint

Placental 11β-HSD2 is thought to protect the fetus from excessive maternal glucocorticoids, which may have deleterious effects on normal growth and development (122). Consequently, in human term fetuses, 75% of cortisol is of fetal origin, whereas cortisone is mainly of maternal origin (123). In addition, fetal tissue 11β-HSD2 may serve to regulate CORT access to GR locally, where the enzyme is expressed widely until midgestation (107, 124). A role for feto-placental 11β-HSD2 in regulating CORT access to MR is unlikely as this receptor type is only expressed in the fetus in the last few days of gestation (106, 107) and barely if at all in the placenta (121).

Placental 11β-HSD activity shows substantial interindividual variation in rodents and humans (88, 125), possibly due to differences in the levels of local or circulating regulators of 11β-HSD activity (126, 127). Interestingly, Montano et al. described greater transport of CORT across the placenta in female than in male mouse fetuses, suggesting that placental 11β-HSD may be less active in female fetuses (128). This may, at least in part, explain why females appear to be more sensitive to prenatal stress than males (32, 45–47).

Interestingly, the activity of placental 11β-HSD correlates positively with birth weight in rats (88) and humans (129), suggesting that relative deficiency of the placental metabolism of glucocorticoids allows maternal levels to access the fetal compartment and thereby attenuate growth. Moreover, humans with deleterious mutations of the 11β-HSD2 gene have very low birth weight (130–132). Inhibition of feto-placental 11β-HSD by administration of carbenoxolone (CBX), a derivative of the active component of liquorice, to pregnant rats decreases birth weight in the offspring, and programmes permanent hypertension and hyperglycaemia later in life (7, 8). These effects require intact maternal adrenals to become manifest (7, 8, 133), suggesting that CBX, by inhibiting placental and perhaps fetal tissue 11β-HSD2, allows the higher maternal levels of CORT to access fetal and placental GR, thereby altering fetal growth and tissue development, hence programming physiological processes in the offspring.

Prenatal CBX administration also increases HPA activity later in life, increasing CRH expression in the PVN, and elevating basal and poststress plasma CORT concentrations (134). Reduced glucocorticoid negative feedback in these animals may underlie these changes, as GR mRNA levels in the PVN are reduced (Fig. 5). Inhibition of feto-placental 11β-HSD also affects behaviour (Fig. 5). Rats prenatally exposed to CBX are consistently less immobile in a forced-swim test. Furthermore, CBX-rats had reduced exploratory behaviour (less rearing) and maintained an increased level of arousal (less grooming) in the open field (134). These behavioural studies show that rats exposed to 11β-HSD inhibitors in utero, and presumably therefore to increased levels of endogenous maternal CORT, exhibit clear alterations in adult behaviour when exposed to stressful environments. These findings are compatible with prenatal stress-induced programming of impaired coping in aversive situations. Anxiety-like behaviour is mediated by the amygdala, probably via CRH (104), the transcription of which is facilitated by corticosteroids (99, 100, 135). Thus, higher GR expression observed in the amygdala of CBX offspring may increase amygdala sensitivity to already elevated CORT levels (134), thereby inducing high levels of CRH that mediates behaviour.

image

Figure 5. Effects of inhibition of feto-placental 11β-HSD. The top graph shows increased CRH (corticotropin-releasing hormone) (black bars) and decreased GR (glucocorticoid receptor) (striped bars) mRNA levels in CBX (carbenoxolone) offspring. The bottom graph shows that CBX offspring spend less time grooming (striped bars) and rearing (black bars) in an open-field test. **P < 0.01, ***P < 0.005. PVN, paraventricular nucleus of the hypothalamus.

Download figure to PowerPoint

DEX is relatively poorly metabolized by 11β-HSD2 (136, 137) and, as detailed above, when administered to pregnant rats alters a range of developmental processes, resulting in growth retardation followed by adult hypertension, glucose intolerance, HPA dysregulation and behavioural alterations (6, 9, 88). These effects are broadly similar to those of prenatal inhibition of 11β-HSD (134). Taken together, these data suggest that feto-placental 11β-HSD2 may play an important role in modulating the programming effects of prenatal endogenous glucocorticoid exposure (138, 139).

Maternal undernutrition

The idea of fetal (or maternal) undernutrition leading to programming of the offspring is based on the rationale that the fetus adapts to poor in utero energy availability by changing the set-point of systems involved in metabolism (1, 3). However, these permanent changes may be counterproductive if nutrition is abundant in the postnatal environment, and may ultimately lead to disease, in other words the ‘thrifty phenotype’ hypothesis. The relevance of this theory for HPA programming may lie in the capacity of glucocorticoids to regulate body weight and food intake (140), as well as perhaps underlying the broader metabolic and cardiovascular phenotypes observed. Of course, such nutritional programming might also affect CNS parameters, such as behaviour and learning, either directly or as a by-product of the changes in glucocorticoid levels.

Food restriction or reduced protein intake during pregnancy reduces birth weight and increases blood pressure and blood glucose and insulin levels in adult offspring (3, 5, 141). Moreover, protein restriction in the pregnant rat programmes elevated CORT levels in adulthood (142). The elevated adult CORT levels contribute to the hypertensive phenotype (143), supporting the notion that HPA programming has wider ramifications than merely an exaggerated response to stress. Only one study has investigated stress-induced behaviour in adult prenatally protein-restricted rats, and found decreased anxiety or reduced behavioural inhibition (144). Interestingly, changes in behavioural suppression are also found in prenatally stressed rats (28). Moreover, stressed pregnant rats have been shown to eat and drink less than unstressed pregnant rats (145).

In an analogous vein, the similarities between the effects of prenatal DEX exposure and prenatal malnutrition or protein restriction, especially peripherally, suggest a common mechanism might underlie these effects. One possibility is that DEX reduces food intake in pregnant rats. Indeed, DEX injected pregnant rats eat less and gain less weight than controls (27, 95). Alternatively, food or protein restriction could activate the dam's HPA axis (140, 146), thereby increasing CORT levels that in turn might affect fetal development, if sufficiently able to bypass placental 11β-HSD2. In rats, HPA activation by undernutrition occurs at least in nonpregnant animals (147). Finally, protein restriction during pregnancy reduces placental 11β-HSD activity, presumably amplifying fetal exposure to maternal glucocorticoids (148). This accords with the observation that the effects of reduced protein intake during pregnancy on offspring blood pressure are dependent on intact maternal adrenals (133). Thus, the prenatal dietary and glucocorticoid exposure models may be linked at several levels.

A caveat to the interpretation of dietary manipulations remains the difficulties of establishing mechanisms and the uncertainties over which components of the diet are responsible for effects seen. Some data suggest that specific fats, which are often used to balance the caloric deficit of diets with selective reduction in protein, may play a determining role, or perhaps the precise proteins involved are key (149).

Downstream mechanisms of glucocorticoid programming

Glucocorticoid exposure plays a central role in prenatal programming. It remains unclear how glucocorticoids affect development, but several scenarios are possible. A key initial issue that is as yet unresolved, is the locus of action of prenatal glucocorticoids to imprint upon offspring CNS development. In principle, glucocorticoids might act on the mother, the placenta and/or the pup. The potent effects of DEX, but not CORT suggest that the fetus (or possibly the trophoblast, the fetal component of the placenta) is the primary target. In support of this assertion, few effects on maternal metabolic or haemodynamic parameters have been shown with DEX, though more studies are needed here (9). Similarly, the clear CNS programming effects of gestational 11β-HSD inhibition again point to the fetus as the key target for maternally derived glucocorticoids. Even if the fetus is the key target we still need to address the mechanisms involved and several have been proposed.

First, the receptors for glucocorticoids, GR and MR, are clearly programmed by prenatal stress/glucocorticoid exposure and are good general candidates for programming throughout the CNS and periphery (150, 151). Glucocorticoids may directly determine transcription of the MR and GR gene in brain and periphery, permanently altering tissue sensitivity to the steroids. Recently, multiple tissue-specific mRNAs for both MR and GR genes have been characterized (78, 152–154). Each alternate mRNA species encodes the same receptor protein, but differs in its 5′-untranslated first exon/promoter site as a consequence of alternate splicing, thus providing distinct regulatory elements for tissue-specific control of expression of MR and GR by, for example, adrenal hormones (152). Time- and tissue-specific activation of these promoters may allow a single early environmental event to imprint upon complex expression patterns of GR and MR. Thus the apparently contradictory effects of late gestational exposure to DEX, which permanently decreases GR mRNA levels in the hippocampus (6), but upregulates GR mRNA in the amygdala and in a specific subregion (periportal) of the liver (9), might reflect distinct effects of DEX upon hepatic, amygdala-and hippocampus-specific promoters. Indeed, in the neonatal handling paradigm, the permanent increase in GR mRNA and protein seen only in the hippocampus, reflects permanent three-fold activation of one hippocampus-specific promoter, without changes in the other five alternate first exons/promoters of the GR gene that are expressed in this tissue (78).

Second, prenatal glucocorticoids permanently influence developing monoaminergic and other neurotransmitter systems (98, 110). These in turn regulate brain MR and GR expression (155–160) as well as a host of other systems (as discussed above). Thus, lesioning the noradrenergic system reduces MR numbers in hypothalamus and amygdala and MR mRNA in hippocampus, but increases GR levels in the hypothalamus, without alterations in hippocampal GR mRNA (155, 160). Disrupting 5-HT neurones reduces both hippocampal GR and MR mRNA levels (158), whereas lesioning the cholinergic system increases them (159).

Third, peripherally, prenatal glucocorticoid exposure affects the regulation of many components of the insulin-like growth factor (IGF) system (161). The IGFs are key to tissue growth and development in utero. Glucocorticoids directly regulate both IGF genes, their myriad binding proteins and their receptors in the fetus and placenta (162, 163). Overexpression of IGFBP-1 reduces growth and programmes hyperglycaemia (164). Interestingly, these transgenic animals have a reduced brain-to-body weight ratio, suggesting that brain development is selectively impaired (164). Additionally, glucocorticoids may affect this critical system indirectly via alteration in maternal and fetal glucose and insulin levels (165, 166). Indeed, prenatal glucocorticoid administration down-regulates the placental glucose transporters GLUT-1 and GLUT-3 (166). Conversely, the IGF system affects HPA function. Thus transgenic mice overexpressing IGF2 in the postnatal period have exaggerated glucocorticoid responses in adulthood (167).

Genetic programming

Whereas in utero programming represents the ‘nurture’ side of programming, the ‘nature’ aspect should not be ignored. It remains an entirely tenable hypothesis that the link between low birth weight and later disease in humans reflects the action of a gene or genes that inter alia determine intrauterine growth and adult physiology. Indeed, some studies have suggested that genes involved in glucose-insulin homeostasis may underlie the epidemiological findings, at least in part (13). However, such classical genetic differences cannot underlie the effects of environmental manipulations to programme adult responses observed in isogenic rodent strains. Clearly environmental effects also occur. That these act upon a genetic background is shown by strain differences in hormonal and behavioural responses to stress in rats and mice (168–170). In addition, different strains are not equally sensitive to perinatal manipulations. Thus, Lewis rats show less birth weight reduction after prenatal stress than the more stress-reactive Fischer F344 strain. In contrast, prenatal stress produces more widespread behavioural adaptations to adult stress in Lewis rats than in the Fischer F344 strain (36). Even substrains of rat differ markedly in their adult phenotypes in response to early life manipulations (171). Similarly, genotype constrains the effects of neonatal manipulations upon adult cognitive function in different mouse strains (172).

An intriguing consequence of these data is that ‘inherited’ components of stress responsiveness may be passed on to future generations either via classical genetic mechanisms or by in utero glucocorticoid/stress exposure. This is likely to be greater in stress-hyperresponsive than hyporesponsive dams, producing transgenerational effects difficult to discern from Mendelian processes. Indeed, cross-fostering studies in mice suggest that perinatal environmental and genetic factors mutually influence each other in determining HPA activity and behaviour later in life (170).

Interactions with the postnatal environment

Postnatal and later environmental events might modulate the effects of prenatal programming. Lifestyle factors such as smoking, exercise and diet clearly influence blood pressure and blood glucose levels. Where brain and HPA axis function are concerned, early postnatal manipulations such as handling (54, 173) and adoption (44) reverse at least some of the effects of prenatal stress. This implies that postnatal manipulations and prenatal stress may have some common neural targets on which they have opposite effects at different time-points. Alternatively, it is possible that the effects of prenatal stress or fetal glucocorticoid exposure extend into the postnatal period and thus can be prevented or modulated by events within that same programming ‘window’.

Maternal behaviour is known to influence offspring HPA activity and anxiety behaviour (174, 175). In fact, the effects on the offspring of neonatal handling are due to increased licking and grooming of pups by the dam, initiated by the short daily separations of the neonatal handling procedure (176). Indeed, the offspring of undisturbed mothers that have a naturally high licking and grooming behaviour are indistinguishable from handled rats (177). Moreover, stroking pups during a maternal deprivation procedure can abolish some of the effects of long-term maternal deprivation (178), whereas maternal treatment with anxiolytics prevents the effects of the handling on offspring emotionality (175). Thus, it is possible that maternal stress and glucocorticoid administration during pregnancy exert their effects through altered maternal behaviour in the early neonatal period, and that environmental manipulations during that period may further modify maternal behaviour. At present, it is not known whether stress or DEX administration during pregnancy alters maternal behaviour, although one paper suggested that at least the duration of anogenital licking of pups does not differ between previously stressed and unstressed mothers (179). Interestingly however, early adoption increases maternal licking and its effects on offspring HPA activity occur irrespective of the dam's experiences during pregnancy (44). These data indicate that the effects of stress on maternal licking behaviour are reversed by cross fostering the offspring at birth.

Additionally, stress or DEX exposure during pregnancy and/or postpartum manipulations could alter the hormonal status of the dam's milk, and this in turn might influence the offspring's HPA axis. Unpredictable stress to dams increases plasma CORT levels for up to 3 weeks after parturition (28), which may be reflected in milk CORT. However, when lactating dams are given CORT through their drinking water, their offspring's basal and stress-induced CORT levels are actually reduced, rendering it unlikely that prenatal stress or DEX effects on plasma CORT levels are indirectly caused by altered CORT levels in milk (180). Nevertheless, the possibility that the adverse programming effects of prenatal life can be beneficially modulated by the postnatal environment remains an intriguing topic of study and may possibly have eventual therapeutic importance.

Implications for humans

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References

Accumulating evidence suggests that low birth weight may be a risk factor for neurodevelopmental disorders such as autism (181) as well as stress-related psychopathologies such as adult depression and perhaps suicide (11). So might the same mechanisms that have been explored in rodents also pertain to humans? Before answering this question, it is important to realize that rats and humans have different rates of brain development, and that caution should be exercised in extrapolating results to humans, particularly with respect to the timing of DEX exposure.

Undernutrition and prenatal stress

In humans, a correlation between maternal nutritional status during pregnancy and childhood blood pressure has been observed (182). Moreover, there is epidemiological evidence from persons conceived during the Dutch Hunger Winter of 1944–45 that prenatal food restriction may, depending on the timing of the exposure, be involved in subsequent development of adult obesity (183), glucose intolerance (184) and even in the origin of (some cases of) schizophrenia (183, 185). Furthermore, early malnutrition reduces brain DNA content (186), lowers IQ scores (187), and affects behaviour and emotionality (188, 189) in children.

The role of stress and glucocorticoids in these effects are unknown. However, a considerable strand of the literature suggests that prenatal stress affects the outcome of gestation. Whilst the data are far from concordant, several studies have found that prepartum maternal stress or depressive illness reduces gestation length, with or without reducing intrauterine growth (adjusted for gestational age) (190–196). Thus, a study of more than 2500 pregnancies in the USA showed that maternal stress, assessed at the beginning of the third trimester, was associated with preterm birth and low birth weight even after adjustment for a host of maternal demographic characteristics and other known risk factors (197).

A number of studies have shown links between maternal stress during pregnancy and psychopathology and disturbed behaviour in children (81, 85, 87, 198). Although it is likely that maternal stress affects the offspring in utero (e.g. altered fetal movement and heart rate (199, 200), it cannot be excluded that mothers stressed during pregnancy have a different relationship with their children after birth. This may also play a role in the changes observed in ‘prenatally stressed’ children. Indeed, early postnatal stress in humans predisposes to developmental delays and behavioural disturbances (86), and possibly also personality disorders (198). This illustrates that events early in life, either preor postnatal, may have long-term consequences, and that birthweight probably merely serves as a marker for these effects rather than being a crucial factor per se in programming.

Glucocorticoid programming in humans

Glucocorticoids are given under two distinct scenarios in obstetric practice. Quite commonly, DEX is given as short-term treatment when preterm labour threatens in the last trimester. This aims to delay labour and to accelerate fetal lung maturation to prevent neonatal respiratory distress syndrome (201). More rarely, DEX is used throughout gestation from the first trimester to attenuate fetal adrenal steroid overproduction in fetuses at risk of congenital adrenal hyperplasia (CAH). Here DEX reduces virilization by attenuating adrenal androgen excess (202). The long-term effects of fetal glucocorticoid exposure in humans have been poorly investigated. A preliminary study in children at risk for CAH who received DEX in early gestation, showed that prenatal DEX exposure did not affect cognition but did decrease scores for psychosocial behaviour (203). Another study reported delayed psychomotor development and a failure to thrive in such patients (204). The effects of late-gestational fetal glucocorticoid exposure on behaviour in humans are unknown, though a very recent report suggests this predisposes to hypertension later in life (205).

Based upon findings in the prenatal DEX-exposed rat model of low birth weight and adult hypercorticosteronaemia, HPA function in adult human birth weight cohorts has recently been investigated. Intriguingly, birth weight correlates closely with HPA measures from infancy (206), through adolescence and young adulthood (207) to old age (208). These data suggest that low birth weight associates with both increased basal and ACTH-stimulated cortisol levels (207, 209). Taken as a whole, these findings are compatible with the hypothesis that fetal overexposure to glucocorticoids, whether exogenous DEX or endogenous cortisol, may underlie at least in part the connection between the prenatal environment and adult stress-related and behavioural disorders. Human neonates exposed to DEX, or endogenous glucocorticoids produced in response to maternal stress during pregnancy, may be at risk of later HPA, behavioural and cognitive disturbance. However, the plasticity of the brain and HPA axis extends far into the postnatal period and animal studies suggest that postnatal events may further modify these genetically and prenatally programmed systems. Such intriguing effects remain to be explored.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References

Work in the authors' laboratory is supported by a Wellcome Trust Programme grant (JRS), a Wellcome Prize Studentship (LAMW) and grants from the Wellcome Trust, Medical Research Council, BBSRC, British Heart Foundation and the European Union.

Accepted 28 August 2000

References

  1. Top of page
  2. Abstract
  3. Programming and the CNS
  4. Prenatal stress
  5. Mediators of early life programming
  6. Implications for humans
  7. Acknowledgements
  8. References