The present study provides evidence for the hypothesis that the extent and the direction of experience-induced synaptic changes in cortical areas correlates with time windows of neuronal as well as endocrine development. Repeated brief exposure to maternal separation prior to the stress hyporesponsive period (SHRP) of the hypothalamic-pituitary-adrenal (HPA) axis induced significantly reduced dendritic spine density (−16%) in layer II/III pyramidal neurons of the anterior cingulate cortex (ACd) of 21-day-old rats, whereas separation after termination of the SHRP resulted in increased spine densities (+16%) in this neuron type. In addition, rats of both groups displayed elevated basal plasma levels of corticosterone at this age. Separation during the SHRP (postnatal days 5–7) did not influence spine density in the ACd, and basal corticosterone levels remained unchanged. In contrast, pyramidal neurons in the somatosensory cortex (SSC) displayed significantly enhanced spine densities (up to 52% increase) independent from the time of separation. These results indicate that alterations in the synaptic balance in limbic and sensory cortical regions in response to early emotional experience are region-specific and related to the maturational stage of endocrine and neuronal systems.
Juvenile stressful experience, such as the separation of a newborn animal from its mother or parents, interferes with the maturation of endocrine and behavioral function in rodents and primates, including man (Heim and Nemeroff, 2001; Pryce and Feldon, 2003), and thereby can affect cognitive as well as emotional capacities throughout life. The observed behavioral alterations are most likely a consequence of synaptic changes within functional neuronal networks. Recent studies in rodents have shown that repeated periods of traumatic experiences during the first weeks of life induce alterations of synaptic connectivities in the limbic anterior cingulate cortex and hippocampus (Helmeke et al., 2001a,b; Ovtscharoff and Braun, 2001; Poeggel et al., 2003).
For sensory systems it has been shown that the functional maturation of cortical circuits is characterized by developmental time windows, during which the impact of sensory experience on synaptic reorganization is particularly strong (Fox, 1994; Katz and Shatz, 1996; Yuste and Sur, 1999; Kral et al., 2001). Immature developing brain regions, in particular, areas of the limbic system, are considered to undergo ‘experience-expectant’ developmental time windows, during which adequate social, emotional, perceptual and cognitive stimulation is required for normal development (Greenough et al., 1987; Joseph, 1999; Andersen, 2003). Such time windows of enhanced synaptic plasticity on one hand serve to fine-tune and adapt the synaptic network according to the individual's environment, but they also represent phases of high vulnerability in cases when no or adverse experiences interfere with the development of brain circuits.
Phases of pronounced synaptic reorganization, i.e. proliferation and elimination of synaptic connections, have also been described for sensory and prefrontal cortical areas of human and non-human primates (Wolff and Missler, 1993; Bourgeois et al., 1994; Huttenlocher and Dabholkar, 1997). However, it is still rather unclear if the described experience driven synaptic alterations, especially in the limbic system, are correlated to distinct developmental time windows and which cellular and endocrine factors define such time windows of cortical development. The aim of the present study was to identify a correlation between the developmental time windows of endocrine stress systems and experience-induced synaptic changes. To test this hypothesis, the impact of repeated periods of emotional stress, evoked by maternal separation, prior to, during or after the stress hyporesponsive period (SHRP) of the hypothalamic-pituitary-adrenal (HPA) axis, on the development of dendritic spines in a limbic and a sensory cortical area was quantitatively analyzed. The SHRP, which is characterized by low basal levels of stress hormones and a relative non-responsiveness to external stressors (Rosenfeld et al., 1992; Levine, 2001), was proposed to protect the juvenile brain against the deteriorating effects of high levels of stress hormones (Meaney et al., 1991). If endocrine function is related to the stress-induced synaptic changes, different changes of spine densities should be expected during these three time windows.
Material and Methods
Male White-Wistar rat pups (Leibniz Institute for Neurobiology, Magdeburg, Germany) were used. The date of birth was designated as postnatal day (PND) 0. After birth, litters were culled to 12 pups (six males and six females) and housed together with their mother in standard rat cages (56 × 34 × 25 cm) in an air-conditioned room with controlled temperature (22 ± 2°C) and a 12/12 h light/dark cycle. Fresh drinking water and rat pellets were available ad libitum. During the entire experiment the cages were not cleaned to avoid handling and stress-related disturbance of the family. All experimental protocols were approved by the ethical committee of the government of the state of Saxony-Anhalt according to the German guidelines for the care and use of animals in laboratory research (§8, Abs. 1, 25.05.1998). All experiments were performed in accordance with the European Communities Council Directive of November 1986 (86/609/EEC).
The pups were removed from their home cage and individually isolated for 1 h in incubators under controlled temperature (35 ± 2°C) and humidity. During isolation no visual or tactile, only auditory and olfactory communication was possible between the siblings. After isolation, the pups were placed back in their home cage.
To test whether separation-induced changes are possibly linked to distinct developmental time-windows, pups were randomly allocated to one of the following four experimental groups (n = 16, four per experimental group, all animals per goup from four different litters).
Naive Control Animals (NC)
These animals were reared undisturbed with their mother.
Maternal Separation prior to SHRP (MS 1–3)
Pups were individually separated from the dam for 1 h per day on PND 1–3.
Maternal Separation during SHRP (MS 5–7)
Pups were individually separated from the dam for 1 h per day on PND 5–7.
Maternal Separation after SHRP (MS 14–16)
Pups were individually separated from the dam for 1 h per day on PND 14–16.
Golgi-Cox Impregnation of Neurons and Quantitative Analysis
After decapitation, the brains were rapidly removed and immersed in 50 ml of Golgi–Cox solution for 14 days. Thereafter brains were dehydrated and embedded in 8% Celloidin. Serial transverse 150 μm sections were collected, mounted on glass slides and processed using a modified Golgi–Cox protocol by Glaser and Van der Loos (1981). Spine densities and dendritic length of layer II/III pyramidal neurons of the anterior cingulate (Fig. 1) and somatosensory cortex were analyzed using the image analysis system NEUROLUCIDA (MicroBrightField, Colchester, VT), which allows quantitative three-dimensional analysis of complete dendritic trees. Only neurons that were impregnated in their entirety and displaying complete dendritic trees were selected. Thirty-two pyramidal neurons per experimental group were analyzed. The length of the dendritic trees was measured by tracing the entire dendrite while counting dendritic spines. However, it is notable that dendritic spines do not represent all synaptic contacts of the analyzed neurons. Mean spine frequencies (number of visible spines per μm) and the length of the dendritic branches were calculated and tested for significant differences between the groups using a Kruskal–Wallis One-Way analysis of variance (ANOVA) followed by a two-tailed Mann–Whitney U-test. All measurements were done by an experimenter blind to the experimental conditions of the animals. An attempt to correct for hidden spines (Feldman and Peters, 1979) was not made, since the use of visible spine counts for comparison between different experimental conditions had been validated previously (Horner and Arbuthnott, 1991).
Stress Treatment and Blood Sampling
Subjects used in this experiment were from different litters than the animals used for the morphological analysis. Blood samples were taken at PND 21 from male pups of the four experimental groups described above applying the following treatments (n = 72):
Basal levels (n = 24, six per experimental group, all animals per group from six different litters): the hormone levels of this group are considered as ‘relative’ baseline values, to which the effect of the conditions described under (2) and (3) can be compared, rather than values of ‘absolute’ resting conditions, since catching the animals per se might act as a stressor. Animals were decapitated within 10 s after removal from the home cage and trunk blood was collected in chilled tubes containing EDTA (BD Vacutainer, Becton Dickinson, UK); samples were centrifuged to separate plasma, which was stored at −80°C until assayed for corticosterone.
Stress challenge (n = 24, six per experimental group, all animals per group from six different litters): the animals were individually placed into an unfamiliar environment (open field arena 50 × 50 × 40 cm) for 30 min, after which the blood samples were taken and handled as described for (1).
Stress recovery (n = 24, six per experimental group, all animals per group from six different litters): after open field experience as described under (2), the animals were returned to their family in the home cage for a recovery period of 45 min, after which the blood samples were taken and handled as described for (1).
Commercially available radioimmunoassys were used for the determination of plasma levels of cortiocosterone (Diagnostic Systems Laboratories Inc., Webster, TX). The intra and inter-assay coefficients of variation for corticosterone were 4.2 and 10.0%, respectively.
Data were analysed by two-way ANOVA with the age at isolation and stress treatment paradigm as experimental factors, followed by post hoc Tukey tests (SigmaStat 2.0, Jandel, Germany), with a level of significance set at P < 0.05. Corticosterone levels had a log normal distribution, and a logarithmic transformation was therefore applied to the data for statistical analysis.
Anterior Cingulate Cortex
The anterior cingulate cortex (ACd) of MS 1–3 rats displayed significantly lower spine frequencies on basal (−17%, P = 0.019) as well as apical dendrites (−16%, P = 0.012) compared with naive control animals (Figs 2a and 3). In contrast, MS 14–16 rats showed higher overall spine frequencies on basal dendrites (+16%, P = 0.05) and no changes on the apical dendrites. In the MS 5–7 rats no effects on spine densities neither on basal nor on apical dendrites were found (Figs 2a and 3).
In the somatosensory cortex (SSC) increased spine frequencies were measured on the basal as well as on the apical dendrites of all three maternal separation groups compared with the controls, irrespective of the time window of separation (Figs 2b and 3). The most pronounced separation-induced spine increase was observed in the MS 5–7 (apical: +52%, basal: +40%, P ≤ 0.001) and MS 14–16 (apical: +48.5%, basal: +38%, P ≤ 0.001) animals, and to a lesser extent in the MS 1–3 group (basal: +17%, P = 0.003).
Experience-induced changes of dendritic length showed no correlation to the time period of maternal separation (Fig. 4), but subtle differences between the two cortical areas were detected. In the ACd the length of apical dendrites was significantly decreased only in MS 1–3 rats (P = 0.005) compared with the naive control animals, whereas no changes were observed in the other two maternal separation groups. In the SSC no changes of apical dendritic lengths were detected. Basal dendrites in the ACd showed significantly increased dendritic lengths in rats separated during (PND 5–7) (P = 0.009) and after (PND 14–16) (P = 0.028) the SHRP, whereas dendritic length remained unchanged in rats separated prior the SHRP. The same tendencies were found in the SSC, where basal dendrites displayed significantly (P ≤ 0.001) elongated dendrites in rats separated during (PND 5–7) and after the SHRP (PND 14–16).
Body and Brain Weights
Measurements of body and brain weights did not show significant differences between the experimental groups.
Measurements of stress hormones revealed an influence of maternal separation on basal stress hormone levels, since significant differences in the basal levels of corticosterone were observed between the four experimental groups. MS1–3 (+71%) and MS 14–16 (+56.8%) rats showed significantly enhanced basal levels of corticosterone at PND 21 compared with naive control animals (P < 0.05), whereas MS 5–7 rats showed corticosterone levels comparable to the naive animals (Fig. 5a). The HPA-axis response to stress challenge and stress recovery was similar in all groups [F(2,71) = 97.4, P < 0.001], i.e. all groups showed significant elevations of corticosterone immediately after maternal separation at PND 21 (stress challenge) (P < 0.05) and a strong reduction after reunification with the family (recovery) (P < 0.05) (Fig. 5b).
Dendritic spines are the main target for excitatory inputs and play a key role in the expression of synaptic plasticity (Koch and Zador, 1993; Harris and Kater, 1994; Segal, 2002). Multiple studies in different systems have shown that these postsynaptic structures are quite sensitive to developmentally, experience or pharmacologically induced synaptic plasticity (Bock and Braun, 1999a,b; Goldin et al., 2001; Segal, 2002; Trachtenberg et al., 2002; Lieshoff and Bischof, 2003). It has been shown in neonatal birds and rodents that positive or negative emotional experiences alter synaptic connectivities in limbic regions (Wallhausser and Scheich, 1987; Bock and Braun, 1998; Helmeke et al., 2001a,b; Poeggel et al., 2003). Thus, quite comparable to the experience-driven maturation of sensory and motor systems (Turner and Greenough, 1985; Rosenzweig and Bennett, 1996), the synaptic development of limbic cortical regions is sensitive towards environmental conditions, in particular emotional experience.
Synaptic Changes in the Prefrontal Cortex Induced by Early Emotional Stress are Dependent on the Time Point of Stress Experience during Early Developmental Phases
The present results clearly demonstrate that the extent and the direction of experience-induced alterations of spine synapses in the limbic ACd and sensory SSC are critically dependent on the developmental time window during which the stressful experience is encountered. Furthermore, our results also reveal region specific differences between the analysed cortical areas. In the ACd spine changes only occurred when maternal separation was applied prior or after the SHRP of the HPA-axis, whereas maternal separation during the SHRP had no influence on dendritic spine densities. Moreover, maternal separation prior or after the SHRP resulted in opposite changes of spine densities. In contrast to the findings in the ACd, pyramidal neurons in the SSC displayed elevated spine densities, irrespective of the time point of separation. The strongest effects were found in the MS 5–7 and MS 14–16 animals and to a lesser extent in the MS 1–3 animals.
Factors that might be reponsible or causally linked to this phase-dependent and cortex–specific spine changes include (i) the maturity and activity of endocrine systems including neuronal mineralo- and glucocorticosterone expression; (ii) the maturity of the sensory systems, which determine the complexity of environmental stimuli perceived by the animal when it is removed from its familiar environment; (iii) the maturity of pre-existent synaptic connections; and (iv) the plasticity potential of the neurons and their synaptic contacts, determined by molecular events.
Correlation with Endocrine Time Windows?
The findings for spine development on layer II/III pyramidal neurons of the ACd are in line with the hypothesis that the time windows for experience-induced synaptic fine-tuning correlate with developmental time periods of endocrine functions. However, although similarily elevated basal corticosterone levels were found in the MS 1–3 and MS 14–16 rats at PND21, spine densities in the ACd changed in opposite directions in these two groups. Thus, the contrasting spine changes might be linked to different developmental profiles of glucocorticoid and mineralocorticoid receptor expression during these two time windows. Prefrontal cortical areas have been shown to be a target for glucocorticoids in rodents and primates (Meaney and Aitken, 1985; Sanchez et al., 2000), where the glucocorticoid receptor mRNA is particularly enriched in the cortical layers II/III and V (Patel et al., 2000). However, the developmental time profile for receptor expression for these layers has not been examined in detail. The expression of glucocorticoid receptors is regulated by early environmental manipulations (Meaney et al., 1996; Avishai-Eliner et al., 1999; Ladd et al., 2004), and corticosterone treatment can induce dendritic reorganization and alterations of spine density in the hippocampus and medial prefrontal cortex of adult rats (Woolley et al., 1990; Wellman, 2001; Seib and Wellman, 2003).
Correlation with Sensory Development?
Following the concept that synaptic development is regulated by the complexity of patterned and converging synchronous neuronal activity, the degree and direction of the synaptic changes in the ACd and the SSC is most likely also determined by preexisting synaptic connections, e.g. the functional maturity of the sensory and modulatory input systems. Animals in the oldest separation group (MS 14–16), ending up with elevated spine densities in the ACd and SSC, were able to detect more complex sensory features associated with the stress situation than the animals from the youngest separation group (MS 1–3), ending up with reduced spine densities in the ACd. The higher spine densities in the ACd of the oldest maternal separation group is in line with the findings in the precocious rodent Octodon degus, whose sensory systems are already functional at birth. In this species elevated spine densities have been found after periodic parental separation (Helmeke et al., 2001a,b; Poeggel et al. 2003). It seems that in the more mature animals the separation from the familiar environment might partly act as ‘enriched environment’, which has been shown to result in elevated synaptic densities (Turner and Greenough, 1985; Rosenzweig and Bennett, 1996). Furthermore, the observed changes of endocrine function and synaptic density after maternal separation may on the one hand have been caused by the separation itself, but on the other hand there is convincing evidence that changes in the quality of maternal care after separation are a critical factor (Levine, 2002; Huot et al., 2004). In addition, the differential, phase-specific changes in the limbic ACd and sensory SSC might be explained by their unimodal (SSC) and associative (ACd) properties. The input into the SSC is mainly confined to somatosensory stimulation, e.g. tactile stimulation during handling, exploration of the unfamiliar environment and maternal behaviors such as grooming and licking. In contrast, the associative properties of the ACd integrate polysensory inputs to achieve much more complex environmental information.
The present study also revealed changes in dendritic length in the ACd as well as in the SSC. In line with findings in the hippocampus of adult rats and primates, where dendritic atrophy has been described in the hippocampal formation and prefrontal cortex after stress exposure (Watanabe et al., 1992; Magarinos et al., 1996; Radley et al., 2004), decreased dendritic length of the apical dendrites of layer II/III pyramidal neurons in the ACd was observed in the youngest separation group (MS 1–3). The other two groups of neonatally stressed rats, however, displayed elongated basal dendrites in both the ACd and the SSC. These differential, phase-dependent differences between apical and basal dendritic lengths, especially in the ACd, are suggestive of a role of differential afferent input activity and different expression of transmitter receptors on the two dendritic arbors.
Alterations in the Stress Hormone Systems are Correlated to the Same Developmental Time Windows as the Experience-induced Synaptic Changes
In addition to the neuronal changes, stress hormone systems are also altered after repeated juvenile stress experience and also depend on specific time windows. Only the MS 1–3 and MS 14–16 rats displayed significantly enhanced basal levels of corticosterone at PND 21, whereas MS 5–7 rats had corticosterone levels comparable to the naive animals. In contrast, acute stress-induced corticosterone elevations during the stress-challenge experiment were similar in all groups at PND 21, which is not in line with studies where a single period of 24 h maternal separation was used as paradigm (Lehmann et al., 2002). In the cited study, 24 h maternal separation during different phases of the SHRP did not induce altered basal stress hormone levels but did result in a significant increase of the corticosterone response to restraint stress in adult animals. These changes were independent of the ontogenetic/SHRP status of the pup.
In human and non-human primates, disrupted mother–infant attachment and premature separation have been considered to impair socio-emotional and cognitive competence and to promote vulnerability to major psychiatric illnesses (Suomi, 1997; Agid et al., 1999; Heim and Nemeroff, 2001). Early separation paradigms in rats and non-human primates provide experimental approaches to identify and characterize the cellular and molecular mechanisms that may underlie the development of mental disorders (Ellenbroek et al., 1998; Nestler et al., 2002). The medial prefrontal cortex is connected to limbic areas such as the nucleus accumbens, the hippocampus and the amygdala (Heidbreder and Groenewegen, 2003), regions which are critical for emotional behavior as well as for learning and memory formation. Moreover, there is evidence that subregions of the prefrontal cortex might be directly involved in the regulation of HPA function (Sullivan and Gratton, 2002), and dysfunctions of HPA-axis regulation have been suggested to be causally related to depression (Holsboer, 2001). Taken together, our results indicate that the synaptic circuits within the limbic ‘reward’ system, which plays a critical role in emotional regulation, learning and memory formation, are adapted to the animal's early postnatal environment and shape the network capacities for species-specific behavioural, socio-emotional and cognitive functions in adolescent and adult animals.
We would like to thank Sabine Westphal and Claus Luley for measurements of blood samples and Aileen Schröter for help with the illustrations. This work was supported by grants from the German Science Foundation (SFB 426) and the Volkswagenstiftung.