The hypothalamic PVN develops to a great extent prenatally, and seems to be part of a functioning HPA-axis from the third trimester onwards [gestational day (GD) 17 in rats], when regional CRH mRNA responses are observed to maternal stress (Fujioka et al., 1999). The observation that CRH synthesis and mRNA expression in the fetal hypothalamus are not yet regulated by corticosteroids until the end of the first postnatal week (Grino et al., 1989; Baram and Schultz, 1992; Yi and Baram, 1993), and local CRHR expression levels are particularly high early in development (Insel et al., 1988), implicates an important role for the stress-induced elevations in CRH signaling mediating the effects of ELS on PVN function.

On the structural level, previous work has indicated that PS increases apoptosis in the fetal hypothalamus (Fujioka et al., 1999; Tobe et al., 2005), but decreases apoptosis in adulthood (Baquedano et al., 2011). Although MS was not found to affect local neuronal density during the SHRP, it increased neuronal density afterwards, which was joined by decreased levels of apoptosis-stimulating proteins and enzymes, whilst cell survival-stimulating protein levels were increased (Irles et al., 2014). These data indicate that ELS influences the structural reorganization of the PVN throughout development, and thereby likely alters its role in HPA-axis regulation.

On the functional level, the effects of ELS on both basal and stress-induced CRH release by the PVN seem to heavily depend on the precise developmental period affected by ELS, the stressor applied, and the age at which the effects are assessed (see Table 1 for an overview of findings). Moreover, the effect of ELS on local CRH signaling might be sex-specific, as PVN CRHR1 mRNA and protein levels were reported to be increased as a consequence of PS in males, but decreased in females (Fan et al., 2009; Wang X. et al., 2013; see Box 3 for an overview of sex-specific modulatory effects of ELS). However, these findings are in contrast with another study reporting no ELS-effects on PVN CRHR1 expression in either sex (Zohar and Weinstock, 2011). As CRHR1-activation in the PVN has been associated with anxiogenic effects (Fan et al., 2013), elevated CRHR1 levels in PS males could underlie the anxious behavioral profile resulting from ELS (Huot et al., 2002; Kalinichev et al., 2002; Daniels et al., 2004; Rees et al., 2006; Aisa et al., 2007; Trujillo et al., 2016). In contrast to potentially increased PVN CRHR1 levels, local CRHR2 expression is unchanged or reduced in both sexes as a consequence of PS (Fan et al., 2009; Zohar and Weinstock, 2011; Wang X. et al., 2013) or MS (Bravo et al., 2011; O'Malley et al., 2011).

In addition to CRH, hypothalamic AVP has subtle stimulating effects on ACTH secretion as well (Gillies et al., 1982) and potentiates the effects of CRH (Giguere and Labrie, 1982; Gillies et al., 1982; Lolait et al., 2007). MS has been found to increase local basal AVP mRNA expression at PND14 (Vázquez et al., 2003), PND21 (Zhang et al., 2012), and PND35 (Veenema and Neumann, 2009), and to elevate local stress-induced AVP mRNA levels at PND6 and PND12 in rats (Dent et al., 2000a), whereas it increases stress-induced fos expression in AVP-positive PVN cells (Zhang et al., 2012). In adults, local stress-induced AVP mRNA and protein levels are higher in both PS (Brunton and Russell, 2010) and MS (Veenema et al., 2006, 2007) offspring compared to controls, although effects might be sex- and stressor-specific (Desbonnet et al., 2008; Brunton and Russell, 2010). Like for CRH, effects of ELS on basal PVN AVP expression are rather heterogeneous. While PS exposure does not affect basal AVP mRNA expression in the male PVN (Lee et al., 2007; Brunton and Russell, 2010), it increases local levels in the females (Bosch et al., 2007; Brunton and Russell, 2010). The number of local AVP-expressing cells has however been found to be decreased due to PS (de Souza et al., 2013) in both sexes. Moreover, effects of PS on AVP expression might be depending on genetic background. Basal PVN AVP mRNA expression in rats bred for low levels of anxiety-related behavior (LAB) were found to be lower compared to rats bred for high levels of anxiety-related behavior (HAB), but PS increased AVP mRNA expression in the LAB rats to levels observed in HAB rats, the latter being not affected by PS (Bosch et al., 2006). Neonatal stress was found to either not affect (Veenema et al., 2006), increase (Veenema et al., 2007; Desbonnet et al., 2008; Murgatroyd et al., 2009; Zhang et al., 2012), or decrease (in females) (Desbonnet et al., 2008) basal expression levels compared to unstressed controls. In one of the studies, increased AVP signaling induced by neonatal stress exposure was associated with a sustained DNA hypomethylation of the Avp gene in the PVN, and turned out to critically mediate the observed hypersecretion of corticosterone and alterations in passive stress coping and memory observed in the offspring (Murgatroyd et al., 2009). However, further research seems necessary to elucidate the exact effects of ELS on AVP signaling.

In the adult PVN, GR mRNA has been localized to cells expressing CRH (Swanson and Simmons, 1989), where GR moderates the glucocorticoid-mediated negative feedback on the HPA-axis by regulating CRH gene expression (Majzoub et al., 1993). Both prenatal exposure to exogenous corticosterone and PS have been shown to decrease local GR expression (Bingham et al., 2013). Inhibition of 11β-HSD2 during pregnancy, raising prenatal corticosterone exposure, also induced reduced GR mRNA expression in the PVN (while it locally increased CRH mRNA levels; Welberg et al., 2000), whereas MS attenuated GR-binding in young rats and decreased GR mRNA levels in adult (Arnett et al., 2015) and senescent animals (Workel et al., 2001). These findings suggest that ELS attenuates HPA-axis regulation at the level of the PVN by reducing GR-mediated negative feedback.

Both corticosteroid and CRH receptors are present in the pituitary from the third trimester onwards (Insel et al., 1988). In contrast to the PVN, pro-opiomelanocortin (POMC; the precursor for ACTH) transcription is already stimulated by CRH and inhibited by corticosteroid administration at this age, implicating functional receptors and local negative feedback regulation well before birth (Scott and Pintar, 1993). However, soon after birth, the pituitary shows a time-limited, reduced response to CRH, which could either be the result of a reduced sensitivity to CRH (Dent et al., 2000b) (although CRHR expression is high at that time) or a reduction in the size and number of ACTH-secreting cells in the pituitary (Sapolsky and Meaney, 1986). Exaggerated negative feedback-sensitivity to corticosteroids (Walker et al., 1986b) might further contribute to this non-responsiveness, but this cannot be readily explained by altered corticosteroid receptor expression levels (which are relatively stable prenatally, and only slowly increase after birth to reach adult levels; Keller-Wood et al., 2006).

Similar to the PVN, the pituitary of PS animals is characterized by decreased cell proliferation and cell death in adulthood (Baquedano et al., 2011). Basal ACTH and POMC expression levels seem to be rather unaffected by ELS in the adult offspring (see Table 2 for an overview of findings). While increased basal ACTH levels have been reported for PS females, ACTH plasma levels seem to be unaffected by neonatal stress. Interestingly, though basal POMC mRNA levels are not influenced by PS (Brunton and Russell, 2010), they are elevated by MS (Murgatroyd et al., 2009), associated with an enduring hypomethylation of the POMC gene (Wu et al., 2014), indicating alterations in ACTH turnover. Concerning stress-induced responses, PS seems to increase the POMC mRNA (in males) and ACTH response to stress (Fan et al., 2009; Brunton and Russell, 2010), at least partially by increasing CRHR expression in the anterior pituitary (Fan et al., 2009). The effects of neonatal stressors on stress-induced ACTH release however seem to again greatly depend on the type of stressor, its timing and duration, and the age of the animal at which the effects are assessed (Table 2). Generally, MS for 24 h both before the onset of and early with in the SHRP is found to increase offspring's ACTH plasma levels in response to stress. However, MS during the second half of the SHRP increases ACTH response to stress only if tested during the SHRP (Smith et al., 1997; van Oers et al., 1998b), whereas it reduces ACTH stress responses measured at an older age (Smith et al., 1997; van Oers et al., 1997, 1998b). AVP seems to play an important role in mediating these effects, as the increase in ACTH levels as a consequence of 24 h MS on PND9 was not observed in AVP deficient animals (Zelena et al., 2015). Since AVP deficiency or AVPR1b antagonist pretreatment diminished ACTH responses to stress only in pups but no longer in adults (Zelena et al., 2011), AVP seems to be particularly important in regulating ACTH-secretion in the neonate. LN seems to reduce ACTH stress-induced responses during the SHRP, though data is limited. Both multiple-day MS and 24 h MS during the SHRP seem to increase ACTH stress responses in adulthood, but not consistently. ESD does not exert any obvious effect (Table 2). Stressor- and age-dependent alterations in local CRHR binding capacity may contribute to the diversity of these effects. ESD and LN are for example found to reduce CRHR binding capacity (Ladd et al., 1996; Avishai-Eliner et al., 2001), and thereby limit the ACTH-releasing potential of CRH.

PS (or prenatal corticosterone) generally increases corticosterone stress responses by elevating peak levels or increasing the total duration of the response (see Table 3), which both appear indicative of impaired negative feedback. Overall, these effects appear slightly stronger in PS females than males (Brunton and Russell, 2010; Table 3, Box 3). Basal corticosterone levels seem to be either increased or unaffected by PS (Table 3).

Effects of neonatal stress on adrenal function are again stressor-specific, and depending on the developmental period affected and the age at which they are assessed (Table 3). The LN model generally induces elevated basal corticosterone levels during the SHRP, which can be prevented by either GR- (in females) or CRHR1- blockage (in both sexes; Liao et al., 2014). However, these levels (as well as adrenal weight) seem to have normalized in adulthood (Naninck et al., 2015), although sex-specific effects might exist; whereas some studies observed increased corticosterone levels and adrenal weight in LN males (Rice et al., 2008; Arp et al., 2016), decreased levels were observed in females (Arp et al., 2016). Corticosterone stress responses have been shown to be either prolonged (Gilles et al., 1996), unaffected (Wang et al., 2012), or reduced (McLaughlin et al., 2016) as a consequence of LN. The effects of 24 h MS seem to be strongly age-dependent as well. MS applied during the SHRP increases both basal and stress-induced corticosterone levels observed during the SHRP (Table 3), without affecting basal corticosterone levels and exerting only minimal effect on stress-induced corticosterone levels when assessed later during infancy. Increased basal levels, but reduced stress-response levels are observed in 3 month-old rats (Workel et al., 2001), whereas in 5 and 12 month-olds basal levels are unaltered, but stress-response levels increased as a consequence of MS (Workel et al., 2001; Lehmann et al., 2002). In elderly rats (20 months), basal and stress-induced levels are again unaffected (Lehmann et al., 2002), whereas stress-induced levels are reduced at senescent age (Workel et al., 2001). Similarly, multiple-day MS does not seem to induce any consistent alterations in basal corticosterone levels either during the SHRP or adulthood (Table 3). Corticosterone stress responses are however typically increased as a consequence of this repeated stressor. Lastly, ESD has been shown to increase basal and stress-induced corticosterone levels during the SHRP (24 h ESD) and in adolescence (PND45) (multiple-day ESD), though in juveniles (PND30) multiple-day ESD was not found to affect basal corticosterone levels, and slowed down stress-induced release. In adulthood, basal corticosterone levels generally are similar to levels observed in non-stressed controls. Interestingly and in contrast to MS, ESD seems to induce reduced stress-response corticosterone levels in adulthood (Table 3).

These ELS-induced alterations in corticosterone plasma levels could obviously be caused by the earlier mentioned alterations in CRH and ACTH release, but could also be attributed to abnormal function of the adrenal gland itself, as increases in adrenal weight and cortex-to-medulla ratio have been reported as a consequence of PS (Ward et al., 2000; Fan et al., 2009; Liaudat et al., 2015). However, the frequent inconsistencies in findings emphasize the extremely complex modulatory effects ELS exerts on the HPA-axis, depending on the precise developmental stage affected, the exact stressor used (its frequency, duration, etc.), age of testing, sex of the offspring, and also the genetic background of the animals. Structured assessment of these effects is absolutely necessary to increase understanding of the underlying mechanisms of aberrant corticosteroid signaling later in life.

The amygdala plays a prominent role in the behavioral fear response and the regulation of emotional processing (Akirav and Maroun, 2007). CRH-expressing cells (first detected at PND6, after which they gradually increase with age; Vazquez et al., 2006) are quite abundant, particularly in the CeA, a major output site which projects to the hypothalamus (LeDoux et al., 1988; Gray et al., 1989). Activation of GRs expressed on CeA CRH-neurons increases local CRH mRNA expression (Makino et al., 1994), which directly contributes to a state of fear (Kolber et al., 2008). These CRH-containing neurons project through the bed nucleus in the stria terminalis to the PVN, and are believed to stimulate the HPA-axis and induce anxiety-like behavior (Feldman et al., 1994; Brunson et al., 2001a). Simultaneously, CRH released by the PVN activates the amygdala to increase anxiety (Schulkin, 2006), forming a potent feed-forward loop in stress signaling.

The amygdala develops both pre- and postnatally. It emerges during the third week of gestation, but matures prominently throughout infancy and adolescence (Berdel et al., 1997), changing neuronal morphology (Ryan et al., 2016), intrinsic membrane properties, action potential kinetics, and the synaptic and voltage-gated currents (Ehrlich et al., 2012, 2013). From PND7–21 in rats, regional soma volume doubles, spine density increases nearly five-fold, whereas dendritic arbors expand throughout the first postnatal month (Ryan et al., 2016). Neuronal density however reduces postnatally (Berdel et al., 1997).

PS influences the developmental trajectories of the rats' amygdalar subnuclei; the BLA, CeA, and lateral (LA) amygdala. Development of these regions was shown to be temporarily impeded by PS, with at offspring displaying significant reductions in regional volume and neuronal and glial number at PND25, which normalized at PND60 (Kraszpulski et al., 2006). In line with this, increased apoptosis was observed in the amygdala of pups (at PND7) as a consequence of prenatal corticosteroid treatment (Zuloaga et al., 2011), and an altered balance in subunit expression of glutamatergic and GABAergic receptors was observed at PND14–22 following prenatal restraint (Laloux et al., 2012). However, in another study the same stressor increased the volume and neuronal and glial number of the LA—the subregion serving as the site of signal-input from the sensory processing systems (LeDoux, 1994)—at PND80–120, without affecting the other subregions (Salm et al., 2004), suggesting age-dependent effects. Cell proliferation in the infant amygdala showed a non-significant reduction as a consequence of PS (Kawamura et al., 2006), whereas electrophysiological recordings from BLA excitatory principal neurons revealed a hyperpolarized resting membrane potential, larger action potential after-hyperpolarizations and H–currents in PS rat offspring compared to controls, reducing neuronal excitability throughout development from infancy into young adulthood (PND60; Ehrlich and Rainnie, 2015).

Whereas PS thus appears to transiently impede amygdala development, stress applied to the neonate seems to hasten amygdala maturation. Typically, the amygdala is not activated by aversive experiences shortly after birth (until PND8) and pups show attenuated learning of fear (and an approach response to aversively conditioned stimuli; Sullivan et al., 2000), which seems to be crucial for forming dam-pup attachment (Sullivan and Holman, 2010). Neonatal stress however accelerates the development of an aversive response and precocious activation of the amygdala, with pups expressing aversive learning and significant corticosterone stress responses at PND8 when reared in the LN model (Moriceau et al., 2006, 2009). This acceleration seems to be mediated by increased corticosteroid exposure, as corticosteroid infusion in the amygdala mimics the effects (Moriceau et al., 2006) and the administration of a corticosteroid receptor antagonist prevents them (Moriceau et al., 2009). In fact, suppressed aversion learning may be another reason for the SHRP, reducing corticosterone exposure to allow proper dam-pup attachment to occur. Neonatal stress also leads to longer fear retention (Callaghan and Richardson, 2012) and precocious expression of the mature form of extinction learning (Callaghan and Richardson, 2011; Cowan et al., 2013); all suggesting a (premature) acceleration in amygdala development of the stressed neonate. Amygdalar connectivity is affected by this early “maturation,” as myelination is expedited due to ELS (Ono et al., 2008). Potentially, this strengthening of early connections (e.g., those to the thalamus and nucleus accumbens) comes at the expense of the connections that form later in development, including those to the frontal cortex (Bouwmeester et al., 2002). Support for this idea comes from the preclinical observation of aberrant functional amygdala-frontal cortex connectivity in adolescents and adults that experienced childhood adversity (Birn et al., 2014; Fan et al., 2014; Lee et al., 2015). Alternatively, these changes in connectivity could derive from the precocious closing of a critical period of plasticity through neonatal stress. Closure of such critical periods has been shown to coincide with the emergence of perineuronal nets on parvalbumin interneurons (Pizzorusso et al., 2002; Hensch, 2005; Dityatev et al., 2007; Nowicka et al., 2009), stabilizing synapses. MS was shown to increase the number of parvalbumin neurons in the periadolescent LA (Giachino et al., 2007; Seidel et al., 2008), but the effects of stress on the perineuronal nets still have to be characterized. Gross amygdala morphology however does not seem to be affected by MS (Krugers et al., 2012).

Functionally, the adult amygdala seems to be come “overactive” as a consequence of ELS. Assessment of regional cerebral blood flow (CBF) by autoradiography revealed an increased cerebral activation of the amygdala in adult (~PND100) PS offspring to a fear-conditioned stimulus (Laviola et al., 2004), which was accompanied by heightened fear responsivity (i.e., freezing behavior; Sadler et al., 2011). However, also increased amygdala and fear responsivity to the tone was observed without any prior conditioning, suggesting general amygdala hyperactivity and increased anxiety in the PS animals (Sadler et al., 2011). In line with elevated amygdala activity, PS or exposure to elevated corticosteroid levels during gestation was shown to increase amygdala's basal CRH mRNA levels (Welberg et al., 2001; Brunton and Russell, 2010), as well as local CRH release in adult animals (Cratty et al., 1995). MS was found to leave local basal CRH mRNA expression unaffected (Bravo et al., 2011), but ESD, a more severe stressor, was shown to increase stress-induced levels (Barna et al., 2003). These findings may be related to local changes in the inhibition of CRH-induced activation as regulated by local GABAergic signaling. GABAa receptor binding was found to be reduced in the CeA and BLA as a consequence of MS (Caldji et al., 2000), joined by an increase in α2/α3 and decrease in α1 subunit mRNA expression; a profile associated with decreased GABA binding (Wilson, 1996). Moreover, these findings might relate to the altered methylation patterns of the Crh promoter as a consequence of ELS, which correlated with CRH mRNA levels in the central amygdala in a learned helplessness paradigm, but their direction depends on the genetic background of the animal (van der Doelen et al., 2015). The influence of ELS on local CRHR expression seems to be age-, sex-, and subregion-specific. PS was found to elevate CRHR1 mRNA expression in the CeA and BLA of males, and in the MeA of females (Brunton et al., 2011), whereas CRHR2 mRNA expression was not affected in the BLA and MeA, but reduced in the basomedial amygdala of males and increased in females (Brunton et al., 2011). MS was found to increase CRHR1 mRNA expression in the MeA during infancy, and decrease CRHR1 and CRHR2 mRNA levels in the CeA (Vázquez et al., 2003). However, in adulthood CeA and BLA CRHR1 mRNA expression levels are actually elevated in MS offspring, and BLA CRHR2 mRNA expression is reduced (Bravo et al., 2011). Importantly, no effects of neonatal stress on CRHR1/2 mRNA expression levels in adulthood are observed when the amygdala is considered as a whole (O'Malley et al., 2011), emphasizing the relevance of studying subregion-specific expression profiles. MS also affects the rather immediate alterations in receptor expression typically observed following acute stress. It attenuates the typical decrease in CRHR1 mRNA expression and raises CRHR2 mRNA levels in response to an acute psychological stressor (O'Malley et al., 2011). As CRHR1 activation by CRH in the amygdala typically serves an activating, anxiogenic role (Dunn and Berridge, 1990; Henckens et al., 2016), elevated expression levels match the overall increase in anxiety-like behavior of ELS animals. In line with this, injection of a CRHR antagonist abolished the increased fear and sensitivity to the environment of the PS offspring (Ward et al., 2000).

ELS also affects corticosteroid signaling in the amygdala. PS was found to increase CeA GR mRNA levels (Brunton and Russell, 2010) and overall GR-binding (McCormick et al., 1995). These effects might be mediated by elevated corticosteroid exposure of the fetus, as GR (but not MR) mRNA levels in the BLA, CeA, and MeA were found to be increased by the inhibition of 11β-HSD2 (Welberg et al., 2000), and BLA MR and GR mRNA expression were increased as a consequence of dexamethasone administration during pregnancy (Welberg et al., 2001). Remarkably, GR expression in the amygdala was found to be reduced in MS offspring, although this effect might be strain-specific. MS reduced amygdala basal GR mRNA expression during the SHRP in C57Bl/6J mice, but not in CD1s (Daskalakis et al., 2014), and this decrease remained present until adulthood (Arnett et al., 2015). Despite the fact that neonatal stress typically induces an anxiogenic phenotype (Huot et al., 2002; Kalinichev et al., 2002; Daniels et al., 2004; Rees et al., 2006; Aisa et al., 2007; Trujillo et al., 2016), this apparent decrease in GR expression was associated with reduced anxiety of the ELS animals compared to controls, which was normalized by lentiviral-mediated restoration of GR levels (Arnett et al., 2015).

The hippocampus, best-known for its role in spatial learning and memory (Block and Schwarz, 1997), plays an important inhibitory role in the regulation of the HPA-axis by its direct and indirect polysynaptic connections to the PVN. Electric stimulation of hippocampal subfields [CA3, dentate gyrus (DG), and subiculum] reduces corticosteroid release (Dunn and Orr, 1984), whereas hippocampal lesions and those of the ventral subiculum increase CRH mRNA levels in the PVN (Herman et al., 1989), and prolong the corticosterone stress response (Herman et al., 1995), respectively. This feedback seems to be relayed to the hypothalamus by indirect projections through the bed nucleus stria terminalis (Herman et al., 2003). Because of its high local GR/MR expression levels, moderate CRHR1/2 levels, and local CRH-expression, the hippocampus is however highly sensitive to the influences of stress (de Kloet et al., 1990; Maras and Baram, 2012). The first 2 postnatal weeks comprise a crucial period in hippocampal maturation (Frotscher and Seress, 2007), as this is when the hippocampal commissural/associational (C/A) pathways establish their synaptic connections on CA3 pyramidal cell dendrites (Bayer, 1980). Disruption of this process can only be partially restored beyond the third postnatal week (Gall and Lynch, 1978), making that stress experienced during this period can profoundly affect hippocampal structure and function.

ELS has been shown to slow the acquisition of spatial learning and/or impair memory under both moderately stressful and relatively stress-free conditions (Lemaire et al., 2000; Huot et al., 2002; Brunson et al., 2005; Ishiwata et al., 2005; Yang et al., 2006b; Aisa et al., 2007; Kosten et al., 2007; Rice et al., 2008; Ivy et al., 2010; Hulshof et al., 2011). In one of these studies, PS-induced learning deficits were associated with a reduction in spine density of pyramidal neuron dendrites in the hippocampal CA3 region (Ishiwata et al., 2005). Other studies confirmed this PS-reduced spine density not only in the CA3, but also the CA1 subregion of the hippocampus (Martínez-Téllez et al., 2009). Besides, PS reduced dendritic length and branching of CA3, but not CA1, neurons (Hosseini-sharifabad and Hadinedoushan, 2007). Similar reductions in spine density of CA1 neurons were observed as a consequence of ESD and LN, which was, in contrast to the case of PS, joined by CA1 dendritic atrophy (Ivy et al., 2010; Monroy et al., 2010). Moreover, LN was found to reduce apical dendritic length and neuronal complexity in CA3 neurons in infants (Liao et al., 2014). Whereas MS decreased the density of mossy fibers in the stratum oriens (Huot et al., 2002), no changes in apical dendritic length and neuronal complexity have been found in the DG (Oomen et al., 2011).

These structural alterations affect local synaptic plasticity; PS impairs long-term potentiation (LTP) in the CA1 (which is associated with a decreased expression and impaired interaction of the NR1 and NR2B subunits of the NMDA receptor in hippocampal synapses; Son et al., 2006), whereas long-term depression (LTD) is facilitated. Furthermore, PS was shown to enhance the effects of acute stress on impairing hippocampal LTP and facilitating LTD (Yang et al., 2006a). Cross-fostering the neonate offspring with control mothers did not change these effects on hippocampal LTP and LTD, implicating they resulted directly from the prenatal manipulation and not altered maternal care (see Box 2; Yang et al., 2006a). However, environmental enrichment after weaning restored plasticity in PS animals, as well as the associated impairments in spatial memory (Yang et al., 2007), emphasizing the impact of the neonate's environment on PS effects. Not surprisingly, disturbed LTP in the CA1, CA3, and DG is also observed as a consequence of stress in the neonate (by both LN and MS; Brunson et al., 2005; Cui et al., 2006; Ivy et al., 2010; Batalha et al., 2013; Cao et al., 2014; Xiong et al., 2014). However, these perturbations are not always found and may depend on the developmental stage affected by stress (Gruss et al., 2008), the sex of the animal (Oomen et al., 2011), and the age of testing (Brunson et al., 2005). Moreover, they might depend on the exact ELS model implemented, since ESD has been found to enhance DG LTP induction and duration in juvenile (Kehoe et al., 1995; Bronzino et al., 1996) and adult (Kehoe and Bronzino, 1999) offspring. Potentially in line with this ESD-boosted hippocampal LTP is the observation that neonatal isolation accelerates the developmental switch in the signaling cascades for local LTP induction (Huang et al., 2005). However, ESD was also shown to prevent acute stress-induced potentiation of LTP in the DG (Wang H. et al., 2013). Future studies should further elucidate the critical dependables in the modulation of the effects of ELS on hippocampal plasticity.

Although several studies have attributed these effects to elevated corticosteroid exposure of the hippocampus (Brunson et al., 2005), suppressing dendritic growth and branching (Alfarez et al., 2009; Liston and Gan, 2011), the presence of both elevated levels of CRH and CRHR1 (with CRHR1 mRNA expression detected at ~300–600% of adult levels at PND6; Avishai-Eliner et al., 1996) during early developmental stages points toward their critical role in development (and thereby particular sensitivity of the brain to their dysregulation). Hippocampal CRH-immunoreactive neurons are already detected at PND1 (Yan et al., 1998; Chen et al., 2001) and numbers increase to peak levels at PND18, after which levels reduce to those observed in adulthood (Chen et al., 2001). Interestingly, at this initial stage of development, hippocampal CRH mRNA is not only detected in basket- and chandelier-type GABAergic interneurons (Yan et al., 1998; Chen et al., 2001) synapsing on somata of hippocampal pyramidal neurons, but also a second population of CRH-expressing neurons is present, possessing the morphology of hippocampal Cajal-Retzius cells. These non-GABAergic neurons disappear by the end of the second postnatal week (Chen et al., 2001), but emphasize the potential modulatory role CRH can have during early development. CRH is tonically released in the hippocampus, as becomes apparent from the abnormal dendritic structure (i.e., hypertrophy), spine morphology, and impaired synaptic potentiation and spatial learning observed when CRHR1s are chronically blocked (Chen et al., 2004) and in mice lacking CRHR1 (Contarino et al., 1999; Schierloh et al., 2007; Wang et al., 2011). However, the balance seems to be critical. CRH applied to slice cultures was shown to reduce spine density (Chen et al., 2008) and induce dendritic atrophy (Lin and Koleske, 2010), whereas CRH administration into the hippocampus recapitulated the learning and memory problems associated with ELS (Brunson et al., 2001b). Importantly, all these effects are observed when corticosteroid levels are maintained at basal levels. Additionally, both CRH mRNA and protein levels are generally upregulated in ELS animals (Wang et al., 2014), the number of CRH expressing interneurons in the CA1 and CA3 is increased (Ivy et al., 2010), and blockage of CRHR1 prevents dendritic atrophy and LTP attenuation, as well as the impairment in memory performance observed in neonatally stressed animals (Ivy et al., 2010). Therefore, elevated CRHR1-activation has been suggested to mediate the ELS effects on hippocampal function (Maras and Baram, 2012); a hypothesis that was further corroborated by the observation that mice lacking CRHR1 are resistant to the detrimental effects of ELS on hippocampal function (Wang et al., 2011).

ELS also affects neurogenesis in the DG, one of the brain's only sites that displays neurogenesis well into adulthood (Drew et al., 2013). Reductions in adult neurogenesis and cell proliferation are observed as a consequence of PS (Lemaire et al., 2000; Morley-Fletcher et al., 2011; Belnoue et al., 2013), with the severity of the reduction depending on the severity of the PS paradigm and gestational stage affected (Mandyam et al., 2008; Madhyastha et al., 2013), with stress later in pregnancy inducing stronger effects. As the DG for the larger part develops postnatally (Altman and Bayer, 1990a,b), this structure may be particularly sensitive to stress during the first weeks of life. In line with this, it was shown that neonatal stress strongly affects DG neurogenesis in a sex-, age-, and possibly species-specific manner. When assessed at the end of the SHRP, neurogenesis was found to be reduced in rats as a consequence of MS (Lajud et al., 2012), but increased in mice exposed to LN (Naninck et al., 2015). At the age of weaning, sex-specific effects were observed following MS, with male rats showing increased neurogenesis, whereas female rats displayed decreased levels (Oomen et al., 2009). Sex-specific effects of ELS were also observed in adulthood, but in an opposite direction; adult neurogenesis was reduced in MS and LN males (Oomen et al., 2010; Lajud et al., 2012; Naninck et al., 2015), whereas no effects were found in females (Oomen et al., 2011; Naninck et al., 2015). For cell death, conflicting results have been found, ranging from unaffected levels in both sexes (Lemaire et al., 2000; Mandyam et al., 2008), to increased levels in PS males (Mandyam et al., 2008).

Potentially related to these effects on neurogenesis and cell survival, volume reductions have been observed in the DG as a consequence of LN (Naninck et al., 2015), but not MS (Huot et al., 2002). ELS is also reported to locally decrease neuron and glia cell numbers (Leventopoulos et al., 2007; Fabricius et al., 2008; Oomen et al., 2011). Other hippocampal regions were not found to be reduced in volume by ELS (Fabricius et al., 2008; Hui et al., 2011; Zalosnik et al., 2014). Cell proliferation seems to be particularly affected in the caudal/ventral part of the DG (Oomen et al., 2010; Hulshof et al., 2011), implying altered hippocampal contribution to emotional behaviors (Bannerman et al., 2004; Fanselow and Dong, 2010) as a consequence of ELS. Alterations in expression levels of the neurotrophic factor BDNF, which stimulates the survival of newborn cells and is involved in cell proliferation, might be mediating these effects on cell proliferation. Hippocampal BDNF levels in female adult offspring were found to be reduced as a consequence of PS, which was related to a decreased DNA methylation in bdnf exon IV. No such effects were however observed in the male offspring (St-Cyr and McGowan, 2015) and another study even reported on increased BDNF levels in PS males (Zuena et al., 2008). Reports on the effects of neonatal stress on BDNF are conflicting, as both increased (Roceri et al., 2004) and decreased BDNF mRNA expression (Kuma et al., 2004) have been observed in MS-exposed infants, and either similar BDNF mRNA (Roceri et al., 2004; Greisen et al., 2005) accompanied by increased BDNF protein levels (Greisen et al., 2005), decreased BDNF mRNA (Aisa et al., 2009), or increased BDNF mRNA levels (Kuma et al., 2004) have been observed in MS adults. Differences in duration and developmental phase affected by the MS paradigm might be responsible for these inconsistencies, although differences in rat strain might contribute as well.

Interestingly, although adult MS animals mostly show normal basal levels of corticosterone (see Table 3), depleting corticosterone (by adrenalectomy) can reverse this suppression of cell proliferation and neurogenesis, implicating inhibited cellular plasticity due to hypersensitivity to corticosterone signaling in the hippocampus (Mirescu et al., 2004). This abnormal sensitivity to corticosterone might be mediated by altered corticosteroid receptor expression or MR/GR balance as a consequence of ELS. PS has been found to decrease hippocampal MR mRNA levels, density, and binding capacity (Henry et al., 1994; Maccari et al., 1995; Koehl et al., 1999; Van Waes et al., 2006; Brunton and Russell, 2010), which could relate to the increased basal CRH levels in the PVN. Moreover, PS was shown to reduce hippocampal GR levels (Henry et al., 1994; Barbazanges et al., 1996; Koehl et al., 1999; Szuran et al., 2000; Chung et al., 2005; Van Waes et al., 2006; Mueller and Bale, 2008; Green et al., 2011; Bingham et al., 2013), attenuating its negative feedback on the HPA-axis, potentially explaining the stronger and prolonged corticosterone responses in PS animals (Chung et al., 2005; Koenig et al., 2005). These effects seemed to be mediated by increased prenatal corticosteroid exposure of the pups, as they were prevented by adrenalectomy in the mothers and reinstated by corticosterone injection in adrenalectomized dams (Barbazanges et al., 1996). MS during the SHRP induced an immediate decrease in CA1 MR (but not GR) mRNA expression in rat pups (Vázquez et al., 1996), whereas MS toward the end of the SHRP induced an immediate downregulation of both CA1 MR and GR mRNA expression (van Oers et al., 1998a). No effects were observed in the other hippocampal subregions in these studies. In adulthood, mixed effects of MS on hippocampal MR mRNA expression are found, with levels found to be either increased in all hippocampal subregions (Ladd et al., 2004), in the DG only (Workel et al., 2001), or unaffected (Ladd et al., 2005; Batalha et al., 2013). GR protein levels are however univocally downregulated in the hippocampus in neonatally stressed adults (Weaver et al., 2004; Aisa et al., 2007, 2008; Batalha et al., 2013; Arnett et al., 2015) although these effects not always translate to the mRNA level (Ladd et al., 2004, 2005; Brunson et al., 2005). Moreover, these effects may be sex-specific and occur only upon repeated stress exposures, as downregulation in GR and MR expression were observed in males, but upregulation of GR was observed in females as a consequence of 24 h MS (Sutanto et al., 1996). Overall, these alterations might result in an increased MR/GR ratio in the hippocampus (Ladd et al., 2004), which may result in an amplified initial stress reaction by increased activation of the membrane MR in a feed-forward fashion, and an impaired containment of this response by reduced membrane and genomic GR-mediated negative feedback (Oitzl et al., 2010).

Studies have recently focused on the putative association between DNA methylation at the GR gene (NR3C1) and ELS, mediating this reduction in GR expression (in males at least). This line of work started with the discovery by Weaver and colleagues that differential levels of maternal care critically modulated methylation levels of the GR promoter exon 1₇, influencing local transcription factor (NGF1-A) binding, histone acetylation, and ultimately hippocampal GR expression and corticosterone responding in the offspring (Weaver et al., 2004). These differences emerged over the first week of life, were reversed by cross-fostering, and persisted into adulthood. Moreover, they were prevented by the central infusion of a histone deacetylase inhibitor, suggesting a causal relation among epigenomic state, GR expression and the maternal effect on stress responses in the offspring. These findings were replicated in a study in human suicide victims with a history of childhood abuse; the hippocampi of early life abuse victims were characterized by decreased GR mRNA levels, GR transcripts of the GR 1F-splice variant, as well as increased methylation of the NR3C1 promoter (McGowan et al., 2009). Another recent study replicated this finding of enhanced DNA methylation at this splice variant and additionally identified altered DNA methylation in other splice variants of the GR promoter (Labonte et al., 2012). Moreover, it showed that this epigenetic response to ELS is brain region-specific, not occurring in the anterior cingulate. Studies like this, as well as the observation that epigenetic mechanisms critically contribute to conferring cell-type identity during development and cell division, suggest that the impact of environmental factors on epigenetic marks is likely to be to some extent cell-type specific, emphasizing the relevance of limiting analysis to appropriate tissues of interest instead of mere analyses of leukocytes (please see McGowan, 2013 for an extensive review on this issue). Nevertheless, these initial human data translate rodent findings to humans, suggesting a common effect of early life environment on the epigenetic regulation of hippocampal GR expression.

The PFC is key to stress coping and emotion regulation (Arnsten, 2009) through its inhibitory connections to both the amygdala (Banks et al., 2007) and the PVN, where it inhibits CRH release (Radley et al., 2006). It represses the HPA-axis predominantly through inhibitory projections from the infralimbic (IL), prelimbic (PL), and anterior cingulate cortex (ACC) that target HPA-axis neurons directly or indirectly (Heidbreder and Groenewegen, 2003), although the exact functional implications for the HPA-axis seem to be subregion-specific (Radley et al., 2006). It represents the functionally most advanced area of the brain with the longest period of maturation. This prolonged development allows for the acquisition of complex cognitive abilities through experience, but also makes it susceptible to factors that can lead to abnormal functioning, which is often manifested in neuropsychiatric disorders (Schubert et al., 2015). Its development starts prenatally with the proliferation and migration of neurons, growth of dendrites, the formation of neural micro- and macro-circuits through efferent/afferent axonal projections, but continues after birth with the initial overproduction of neurons and their connection being fine-tuned by reducing synaptic contacts (e.g., by the pruning and cell death of unused connections; Kolb et al., 2012) and neuronal density steered by experience.

ELS typically impairs PFC function in adulthood, as is exemplified by increased impulsivity (Gondré-Lewis et al., 2016), deficits in extradimensional shifts of attention (Mehta and Schmauss, 2011), and impaired working, short-term, and long-term memory (Gué et al., 2004; Markham et al., 2010; Negrón-Oyarzo et al., 2015; Alteba et al., 2016). In line with this, PS has been shown to impair prefrontal LTP, which was accompanied by an increase in the mean frequency of spontaneous excitatory postsynaptic currents (sEPSCs) in layer II/III pyramidal neurons (Sowa et al., 2015). Similar results have been observed in rats following chronic corticosterone treatment (Bartosz et al., 2011), suggesting a role for glucocorticoids in this impaired LTP. MS has been shown to result in LTP impairment in the IL layer II/III-layer V (Xiong et al., 2014), and ELS to impair extinction retrieval of context-dependent fear memories by preventing the synaptic potentiation of hippocampal-PL cortex neural pathway, which displayed synaptic inhibition rather than potentiation (Judo et al., 2010). Another study into the effects of PS confirmed this aberrant hippocampal-PFC functional connectivity as the temporal coupling between neuronal discharge in the medial PFC (mPFC) and hippocampal sharp-wave ripples was decreased by PS (Negrón-Oyarzo et al., 2015). In line with this, decreased regional CBF was elicited in the dorsal mid-cingulate and posterior cingulate cortex in MS rats in response to a conditioned tone compared to controls (Sadler et al., 2011).

These functional changes in the frontal cortex might be mediated by structural alterations caused by ELS. Comprehensive insight into the modulatory role of ELS on PFC neural morphology is derived from a series of experiments in a precocious rodent, the degu. In contrast to classical laboratory rats and mice, degus (like human babies) are born with relatively mature sensory systems and can thus perceive and more elaborately interact with their early life environment, making them a very suitable model to study the impact of neonatal stress (Bock and Braun, 2011; Braun and Bock, 2011). Brief MS increases corticosterone levels (Gruss et al., 2006) and both MS and ESD downregulate PFC activity during the separation period (Bock et al., 2012). Repeated separation has been shown to increase spine density in the basal dendrites of layer III dorsal ACC neurons in adolescent animals when compared to nonstressed controls (Helmeke et al., 2001). This finding could potentially be explained by either delayed or permanently impaired synaptic pruning during PFC development. The effect of this increased excitatory spine density may even be exaggerated by a decrease of inhibitory shaft synapses on the neurons by stress (Ovtscharoff and Braun, 2001), inducing a dysbalance of PFC synaptic input and neuronal output. Besides altering synaptic contacts, neonatal stress was shown to also affect the number and type of inhibitory interneurons in the ACC (Helmeke et al., 2008), as wells as reduce mPFC GABAa receptor binding (Caldji et al., 2000), further substantiating evidence for a transient dysbalance in small neuronal feedback loops, and potentially providing a substrate for the development of dysfunctional large-scale neuronal networks. Work in the classical rodent models has substantiated these findings of altered PFC development by ELS, reporting on alterations in both dendritic length and regional spine density depending on the age (developmental stage) and the molecular layer in which they are assessed. Mild PS was found to increase spine density in layer III cingulate cortex neurons at weaning (Mychasiuk et al., 2012), and a mild postnatal stressor caused similar effects assessed pre-puberty (PND35; Monroy et al., 2010). However, in adulthood, PS was found to reduce spine density and dendritic branching and length in dACC and orbitofrontal cortex layer II/III pyramidal neurons (Murmu et al., 2006), and to reduce the ratio of mushroom spines; the type forming the most strong and stable synapses (Michelsen et al., 2007). ESD was also found to reduce apical dendritic length in several PFC subregions in the adult offspring, and reduce spine density in frontal cortex layer III neurons (Monroy et al., 2010; Romano-López et al., 2016). Reduced local expression of BDNF mRNA in adult MS (Roceri et al., 2004), LN (Roth et al., 2009), and PS offspring due to increased Bdnf DNA methylation [associated with an increase in DNA methyltransferase 1 (DNMT1) expression; Roth et al., 2009; Dong et al., 2015], may relate to these changes. However, findings are not indisputable (Muhammad et al., 2012; Boersma et al., 2014), and it has been suggested that the extent and direction of the effects of ELS on frontal neuronal morphology may depend on the developmental status of the neuronal layer at the time the stress is experienced. ESD on PND1-3 was shown to decrease dendritic spine density in layer II/III neurons of the ACC, but failed to have an effect when applied on PND5-7. ESD on PND14-16 however increased spine density on these neurons (Bock et al., 2005). Conversely, ESD on PND5-7 reduced spine density on layer V pyramidal neurons, whereas ESD during the other time intervals did not induce any effects (Gos et al., 2008). As pyramidal cells in layers V/VI are ontogenetically older and therefore establish their synaptic connections earlier than layer II/III pyramidal neurons (Zhang, 2004), these neuron-specific responses to ELS may be due to their different degree of maturity at the time the stressor is experienced. The differential innervation and receptor patterns amongst neuronal layers may be an alternative explanation for their differential sensitivity toward stress (e.g., Zilles et al., 1993). Furthermore, the effects of stress exposure might critically depend on its intensity. Supporting this idea, stressor intensity was found to critically modulate frontal cortex global methylation levels; mild PS increased overall methylation levels at PND21, whereas intense PS induced the opposite effect (Mychasiuk et al., 2011).

Prefrontal CRHR1s have been associated with anxiety (Sotnikov et al., 2014), contribute to the HPA-axis stress response (Jaferi and Bhatnagar, 2007) and were recently found to mediate acute stress-induced executive dysfunction (Uribe-Marino et al., 2016). MS has been observed to significantly increase CRHR1 protein levels in response to acute stress, compared to non-MS controls (O'Malley et al., 2011). However, reported effects of MS on basal CRHR1 expression levels have been mixed, with both no effects (O'Malley et al., 2011) and decreases reported (Ladd et al., 2005). Importantly, ELS reduces GR expression in the PFC, and thereby compromises its negative feedback function on the HPA-axis (Diorio et al., 1993). PS was shown to reduce PFC GR protein levels (Green et al., 2011; Bingham et al., 2013) and binding capacity (McCormick et al., 1995). Also LN and MS induce a significant reduction in PFC GR density (Avishai-Eliner et al., 2001; Ladd et al., 2004, 2005), although this effect is not consistently observed (Huot et al., 2004). Interestingly, a study in which monkeys were prenatally treated with the synthetic glucocorticoid dexamethasone did not observe a decrease in GR expression (Heijtz et al., 2010), suggesting that increased corticosterone exposure in itself is not sufficient to establish these effects. Instead, findings in MS pups of which the dam was given a foster nest for the duration of the separation period implicated a critical role for maternal stress (and potentially care) in influencing GR expression; MS pups from a dam with a foster nest showed increased instead of decreased GR density, accompanied by an (albeit partial) restoration of PVN CRH mRNA levels and ACTH response to stress (Huot et al., 2004). Future research should investigate the exact aspects of the maternal behavior that mediate these normative effects.