The SR mouse model consists of three independent mouse lines selectively bred for either high (HR) or low (LR) HPA axis reactivity in response to a psychological stressor. The IR mouse line, bred for intermediate stress reactivity, serves as a reference line. Briefly, to generate this animal model, a founder generation of 100 outbred CD-1 male and female mice was tested in the stress reactivity test (SRT), which measures the animal’s CORT release in response to a psychological stressor (15 min restraint, a detailed description of the SRT is provided in section “Stress Reactivity and CORT Measurement”). Based on the test results, breeding pairs were selected to generate the HR, IR, and LR mouse lines. Through repeated testing and re-selection of every new generation at the age of 7–8 weeks, three inbred mouse lines were established. The SR mouse model has been extensively phenotyped and several parallels regarding symptoms associated with MDD have been highlighted. In the HR line, these include a reduced bodyweight, increased locomotor activity, hyperactive stress-coping behavior, altered sleep architecture with increased rapid eye movement (REM) sleep, and impaired cognition, akin to endophenotypes of melancholic/psychotic depression. On the other hand, LR animals show an increased bodyweight, reduced locomotor activity, passive stress-coping behavior, and intact sleep and cognitive function, in line with endophenotypes of the atypical depression subtype (Touma et al., 2008, 2009; Knapman et al., 2010a,b, 2012; Fenzl et al., 2011; Pillai et al., 2012; Heinzmann et al., 2014).

To induce ELS, we used the limited nesting and bedding material paradigm (Rice et al., 2008), which creates a stressful early-life situation for the dam and her pups, without having to physically remove the dam from the litter. This ELS paradigm has been described as more ecologically valid than maternal separation (Molet et al., 2014) and its lasting effect on the offspring has been replicated in rats and mice in several studies (Avishai-Eliner et al., 2008; Gunn et al., 2013; Machado et al., 2013; Naninck et al., 2015; McIlwrick et al., 2016). Briefly, 16 dams of each SR mouse line were randomly assigned to either the ELS or the standard (STD) housing condition. On P2, after regulating the litters (see Breeding of Experimental Animals), dams in the ELS condition were placed, together with their pups, into a macrolone cage type II, the floor of which was covered by an aluminum grid (mesh dimensions 0.4 cm × 0.9 cm, catalog no. 57398; McNichols Co., Tampa, FL, USA). Twenty grams of wood chip bedding and half a nestlet (∼5 g) were provided for nest building. Dams in the STD condition were moved, together with their pups, into standard macrolone cages type II with ample wood chip bedding (∼100 g) and 2 nestlets (∼20 g) for nest building. The animals were then left undisturbed until P9, when all litters were moved to standard cages, where they stayed with their mother until weaning (P25). In a previous study using this ELS paradigm in the SR mouse model, we conducted a detailed analysis of the maternal behavior (McIlwrick et al., 2016). Briefly, the dams of all three mouse lines show similar changes in their maternal behavior (number of exits and time spent on the nest) when rearing their pups in the ELS condition. Emerging differences between the pups can thus not be attributed to differences in maternal care between the three mouse lines.

To investigate the gene x environment interaction of a genetic predisposition for extremes in stress reactivity with ELS, we used a three-by-two experimental design (i.e., the three SR mouse lines, HR, IR and LR, and two conditions, ELS and STD), thus resulting in a total of six experimental groups. This two-factorial design was employed both when testing animals during early and late adulthood. In total, the data presented in this manuscript was collected from three sequential cohorts of experimental animals, generated from breeding generations XXIII, XXVI, and XXVII of the SR mouse model. Animals from experimental cohorts I and II were tested starting at 16 ± 1 weeks of age (early adulthood), while animals from cohort III were tested starting at 26 ± 1 weeks of age (late adulthood). There was some natural stratification in the birth dates of the litters of each breeding cohort (∼14 days), so that the mean age of the animals in each cohort was used to determine the date to start testing. An overview and timeline of the experiments included in this study, and of the number of animals used, is provided in Supplementary Figure S1.

For this study, male animals from the three SR mouse lines (HR, IR, and LR) were used. All mice were bred in-house, and housed in sibling-pairs until 2 weeks before the behavioral testing started, when they were single housed to avoid influences of dominance hierarchy. Mice were housed in macrolone cages under standard laboratory conditions, with standard chow and water ad libitum, a 12 h light/dark cycle (lights on at 8:00) and constant humidity (55 ± 10%). Cages were changed once per week, but never on the day of or immediately before testing.

During and after the ELS paradigm, the animals’ bodyweight was closely monitored. Research has shown that bodyweight can be lastingly affected by stress manipulations during early-life (reviewed in Maniam et al., 2014), making this readout a valuable indicator of the impact of the ELS paradigm. The mean pup weight per litter was assessed on P2, P9, P17, and P25, and in adulthood bodyweight was measured 2 weeks before the behavioral testing.

All tests were performed between 08:00 and 13:00 h, when the animals’ CORT levels are in the circadian trough (Ishida et al., 2005; Touma et al., 2009). Between the different behavioral tests, the animals were allowed at least 48 h of rest in order to avoid carry-over effects influencing behavioral readouts in subsequent tests (McIlwain et al., 2001). During the tests, the animals’ behavior was video-recorded and tracked using automated tracking software (ANY-maze, Stoelting GmbH), with the exception of the water cross-maze test, which was scored in real-time by a trained observer. At the beginning and between different animals, the testing apparatus was cleaned with soapy water and 10% ethanol solution and dried to remove any odor cues.

As a measure of HPA axis responsiveness we used the SRT, as previously described (Touma et al., 2008). Briefly, each mouse was removed from its home cage and an initial blood sample was obtained through a small incision from the ventral tail vessel (to ensure the reference sample was very close to baseline levels, the time between initial handling of the cage until completing the blood sampling was less than 2 min). The mouse was then placed into a small restrainer (50 ml plastic tube, with holes for ventilation and an aperture in the cap for the tail) for 15 min, whereafter it was decapitated (after a very brief isoflurane anesthesia), and a “reaction sample” was collected from the trunk blood. We employed the SRT to measure the animals’ stress reactivity during both early and late adulthood (cohorts II and III). All blood samples were kept on ice until centrifuged (4°C) and plasma was removed for measurement of CORT using radioimmunoassay, according to the manufacturer’s protocol (DRG Instruments GmbH, Marburg, Germany), with slight modifications (for details see Touma et al., 2008). All samples were measured in duplicates and the intra- and inter-assay coefficients of variation were both below 10%.

To detect lasting consequences of ELS exposure at the level of gene regulation in the three SR mouse lines, we analyzed the relative expression of selected candidate genes using quantitative polymerase chain reaction (qPCR). Early and late adulthood samples were collected from cohorts I and III to be able to see if the effects of ELS changed over time. Briefly, after the SRT on the last day of testing, the animals were decapitated, the brain was removed, snap frozen in iced methylbutane, and stored at -80°C until further processing. The brains were sectioned into 200 μm thick coronal slices and mounted onto glass slides. Tissue punches of the dorsal (-1.2 to -2,0 mm from Bregma) and ventral (-3.0 to -3,80 mm from Bregma) hippocampus (dHip and vHip) were collected via micropuncture (for further details see Heinzmann et al., 2014). Total mRNA was extracted using RNeasy columns (RNeasy Micro Kit, Qiagen, Hilden, Germany) and 200 ng of the extracted mRNA was reverse transcribed to cDNA using high-capacity transcription kits (High-Capacity cDNA Reverse Transcription Kit, Applied Biosystems, Foster City, CA, USA). The expression of candidate genes was measured using qPCR kits (QuantiFast SYBR Green, Qiagen GmbH, Hilden, Germany) following the manufacturer’s protocol. All samples were measured in duplicates [Standard deviation (SD) < 1.0] on 384 well-plates with three genes per plate, including standard curves. A list of all measured candidate genes with the applied oligonucleotide primers is provided in Table 1. The relative fold expression of each gene was calculated using the ΔΔCT method (Livak and Schmittgen, 2001) by normalizing to two housekeeping genes [TATA-binding protein (Tbp) and Hypoxanthine-Guanine Phosphoribosyltransferase (Hprt)] and again normalizing to the mean of IR STD group.

All statistical data analysis was conducted in PASW 18 or in Python. We used a two-way analysis of variance (ANOVA) with the independent variables “line” and “condition” to assess the main effects and the interaction of these factors. To detect the difference between early and late adulthood data, “age” was added as an independent variable. When data points were collected repeatedly from the same animal in one test, a three-way repeated-measures ANOVA, with “line” and “condition” as between-subjects factors, and “sampling time” as a within-subjects variable, was employed. Where appropriate, post hoc tests were conducted and corrected using the Bonferroni method. The association between cognitive performance and gene expression was investigated using the Pearson’s correlation coefficient. Statistical significance was accepted for ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001, while p ≤ 0.1 (T) was considered a trend.

Some data from cohort I animals was presented in a previous publication (McIlwrick et al., 2016) (OFT, expression of hippocampal Crh and Crhr1) and is shown here again, as we aimed to provide both an early and a late adulthood assessment for all read-outs, without increasing the number of animals sacrificed for this research.

Figure 1A illustrates the development of the animals’ bodyweight throughout the entire experimental time span. Before the start of the ELS paradigm, on P2, there were no significant differences in bodyweight between pups of the three mouse lines, or between pups assigned to the ELS or STD condition (Figure 1B). After 1 week of ELS or STD housing, on P9, the analysis revealed a significant main effect of condition (F_1,90 = 71.557, p < 0.001, post hoc tests: all p < 0.001), showing that pups that had been exposed to ELS had gained significantly less weight than STD-housed pups (Figure 1C). In addition, there was a small difference in bodyweight between the mouse lines (F_2,90 = 2.846, p = 0.063, post hoc tests: HR vs. LR: p = 0.089, all other between-lines comparisons: p > 0.1). On P17, the main effect of housing condition remained significant (F_1,87 = 20.396, p < 0.001, post hoc tests for ELS vs. STD: HR: p = 0.004, IR: p = 0.012, LR: p = 0.023) and a comparison between the three lines revealed that LR pups weighed more than HR pups (F_2,87 = 3.339, p = 0.040, post hoc tests: HR vs. LR: p = 0.036, all other: p > 0.1) (Figure 1D). At weaning on P25, only the main effect of condition was significant (F_1,87 = 18.181, p < 0.001, post hoc tests: HR: p = 0.032, IR: p = 0.007, LR: p = 0.016) (Figure 1E). During early adulthood, there was a clear differences in bodyweight between animals of the three lines, increasing from HR to IR to LR (F_2,127 = 169.490, p < 0.001, post hoc tests: all p < 0.001), as well as a main effect of ELS exposure (F_1,127 = 17.543, p < 0.001), and an interaction of line and condition (F_2,127 = 5.317, p = 0.007). Further analysis specified that the ELS effect was significant in the HR (p < 0.001) and the LR mouse line (p = 0.026) (Figure 1F). During late adulthood, HR mice still weighed significantly less than animals of the other two lines (F_2,65 = 54.141, p < 0.001, post hoc tests: HR vs. IR and LR: p < 0.001), but IR and LR mice no longer differed (IR vs. LR: p = 1.0). In addition, the effect of ELS housing was still significant (F_1,65 = 5.331, p = 0.024), and post hoc tests showed a statistical trend in the LR line (p = 0.096) (Figure 1G).

The absolute levels of plasma CORT concentration before (initial) and after (response) 15 min of restraint were analyzed using a repeated-measures ANOVA. In early adulthood, the data showed a strong effect of time point (initial vs. response), confirming a significant rise in plasma CORT levels in response to the stressor in animals of all three lines (within-subjects effect: F_1,54 = 1459.996, p < 0.001) (Figure 6A). The initial CORT levels (taken within 2 min after the first disturbance of the animals’ cage) showed that HR and IR mice had higher baseline CORT levels than LR mice (F_2,54 = 6.253, p = 0.004, post hoc tests: HR vs. IR: p = 1.0, HR vs. LR: p = 0.007, IR vs. LR: p = 0.015), but there was no difference between conditions (F_1,54 = 0.148, p = 0.702). In the response levels measured after restraint, the main effect of mouse line was exacerbated, with evident differences in CORT concentrations between all three lines (F_2,54 = 214.164, p < 0.001, post hoc tests: all p < 0.001). In addition, the analysis revealed a main effect of condition (F_1,54 = 5.675, p = 0.021), as well as an interaction of line and condition (F_2,54 = 4.232, p = 0.020), and post hoc comparisons showed that HR ELS mice had significantly higher plasma CORT levels in response to the stressor than HR STD mice (p < 0.001), while no other mouse line showed this effect of ELS.

As there had been differences between the groups at baseline, we also analyzed the CORT increase (reaction CORT minus initial CORT level) and this measure confirmed the main effect of mouse line (F_2,54 = 223.319, p < 0.001, post hoc tests: all p < 0.001) (Figure 6B). In addition, the increase in plasma CORT concentration also revealed a main effect of condition (F_1,54 = 6.022, p = 0.017), as well as an interaction of line and condition (F_2,54 = 6.243, p = 0.016), showing that HR ELS mice had a significantly higher increase in CORT levels than HR STD mice (p < 0.001), while IR and LR mice did not show this effect of ELS exposure on stress reactivity (p = 0.776 and p = 0.949, respectively).

The SRT carried out in mice during late adulthood matched our earlier results (Figure 6C). Again, there was a strong effect of time point (initial vs. response) (F_1,53 = 1957.027, p < 0.001) and the initial CORT values showed a strong trend for a main effect of line (F_2,53 = 3.023, p = 0.057, post hoc tests: HR vs. IR: p = 0.063, HR vs. LR: p = 0.309 IR vs. LR: p = 1.0), but no effect of condition (F_1,53 = 0.381, p = 0.540). The difference between the three lines became highly significant after 15 min of restraint (F_2,53 = 140.272, p < 0.001, post hoc tests: all p < 0.001), when, in addition, there was also an interaction of line and condition (F_2,53 = 3.468, p = 0.038). Specifically, ELS-exposed HR mice showed a significantly stronger CORT response than STD-raised HR mice (p = 0.048), while IR and LR mice seemed more resilient to this early-life programming of stress reactivity. The analysis of the increase in plasma CORT levels confirmed a main effect of mouse line (F_2,53 = 166.189, p < 0.001, post hoc tests: all p < 0.001), and the interaction of line and condition (F_2,53 = 4.544, p = 0.015, post hoc tests: HR: p = 0.026, IR: p = 0.092, LR: p = 0.344) (Figure 6D).

The relationship between the animals’ cognitive test performance in late adulthood and the expression of Bdnf, Crh, and Crhr1 in their dorsal and ventral hippocampus was investigated using the Pearson’s correlation coefficient. Our analyses revealed a significant positive correlation between Bdnf expression in the dorsal hippocampus and cognitive performance in the object recognition test (r = 0.265, p = 0.048). In addition, dorsal hippocampal Bdnf expression showed a strong statistical trend for a positive association with the mean performance in both cognitive tests [(discrimination ratio in the Y-maze test + discrimination ratio in the Object recognition test)/2] (r = 0.258, p = 0.059). On the other hand, the expression of Crh in the dorsal hippocampus showed a significant negative correlation with cognitive performance in the Y-maze test (r = 0.329, p = 0.017).

Exposure to the ELS paradigm caused a substantial delay in bodyweight development in pups of all three SR mouse lines, evidenced by reduced bodyweight gain from P2 to P9 in litters raised in ELS conditions (Figure 1C). This main effect of ELS reflects a pronounced impact of the limited nesting and bedding material paradigm on the animals’ physiology and confirms previous findings in the SR mouse model, as well as in other rodent models (Gilles et al., 1996; Avishai-Eliner et al., 2008; Rice et al., 2008; Naninck et al., 2015; Bath et al., 2016; McIlwrick et al., 2016). The difference in bodyweight remained significant all throughout development into early and late adulthood in the HR and the LR lines, similar to observations by others (Bath et al., 2016). Animals in the IR ELS group matched their STD housed control group by early adulthood (Figures 1D–G). However, in a previous study in our animal model (McIlwrick et al., 2016), and as reported by others (Rice et al., 2008), no weight differences were observed between ELS and STD mice by early adulthood. The lasting effects of ELS on bodyweight will need to be investigated in future studies to allow better understanding of the circumstances under which long-term metabolic changes occur.

In line with previous studies using the limited nesting material paradigm or maternal separation (Brunson et al., 2005; Millstein and Holmes, 2007; Rice et al., 2008), ELS did not affect anxiety-related behavior in the adult offspring (Figures 2C–F). In general, the OFT results showed that HR mice were more active than LR mice, a phenotype which was present at both measurement time points (Figures 2A,B) and confirms earlier studies (Touma et al., 2008; Heinzmann et al., 2014). While there was an indication for reduced locomotor activity in aged HR ELS mice (Figure 2B), this finding was not confirmed in a slightly different testing set-up (Figure 4D), pointing toward a context-specific, rather than a general effect of ELS.

Cognitive function was clearly influenced by ELS in the SR mouse model. However, the three lines were not equally vulnerable to the deleterious effects of early-life adversity. HR ELS animals showed the most pronounced phenotype in terms of cognitive impairments, emerging in early adulthood [indications for reduced spatial learning performance (Figure 3) and significantly impaired place memory (Figure 4A)] and lasting into older age [impaired place memory (Figure 4B) and impaired object memory (Figure 5A)]. IR ELS animals showed a similar effect in early adulthood [partially reduced spatial learning (Figure 3) and impaired place memory (Figure 4A)], but appeared to recover with increasing age [intact place memory (Figure 4B) and no significant difference in object memory (Figure 5A)]. In contrast, LR ELS animals showed no indication of cognitive deficits in early adulthood [normal spatial learning (Figure 3) and good place memory (Figure 4A)], and there was even some evidence for improved cognitive function in aged ELS-exposed LR mice compared to STD animals (Figure 4B). These results highlight that the consequences of ELS on cognitive function depend very much on the inherited predisposition of the individual. Previous studies in the SR mouse model have provided evidence for deficits in cognitive function in HR animals (Knapman et al., 2010a,b) and have linked this to reduced hippocampal activity and neuronal integrity (Knapman et al., 2012). The proposed mechanism underlying this phenotype in HR mice is a cumulative neurotoxic effect of glucocorticoids, as lifetime exposure to elevated stress hormones can give rise to progressive deficits in learning and memory (Hibberd et al., 2000; Sapolsky et al., 2000; Lupien et al., 2009). Our new data complements these findings by showing that the cognitive deficits observed in HR mice become exacerbated by ELS exposure. In a set of studies, Brunson et al. demonstrated that ELS can set off a cascade of structural and functional changes in different subfields of the hippocampus, including aberrant mossy fiber expansion, impaired long-term potentiation (LTP) and dendritic atrophy, which contribute to several cognitive impairments emerging with increased age (Brunson et al., 2005). We suggest that these same mechanisms may be acting in HR mice and that they may become increasingly detrimental through the additional sensitization of the HPA axis set in motion by the exposure to ELS.

To verify the central role of HPA axis programming in the effects of ELS, the animals’ stress reactivity was tested at two time points. Both in early and in late adulthood, the results showed that HR ELS mice had an increased stress reactivity compared to STD-raised HR mice (Figure 6). This ELS-induced augmentation in CORT release supports the hypothesis that the cognitive deficits displayed by HR ELS animals are due to excessive, cumulative glucocorticoid exposure and its adverse downstream consequences in stress sensitive regions of the brain, such as the hippocampus. Earlier work from our group demonstrated that directly after a week-long period of ELS exposure HR ELS pups had elevated basal CORT levels, which normalized by the age of weaning and remained low in adulthood (McIlwrick et al., 2016). In the present study, we again assessed the CORT levels in adult animals and found no differences between HR ELS and STD mice at baseline. The stress-induced CORT levels, however, did show a significant impact of ELS, replicating our previous results. This gradual shift from elevated basal CORT levels in pups to enhanced stress reactivity in adulthood suggests that a disruption of the HPA axis suppression during the stress-hyporesponsive period (SHRP) led to changes in the neuroendocrine programming of stress reactivity in these animals. During the SHRP, lasting from P2–P12 in mice, moderate stressors fail to elicit a measurable physiological stress response in the pups, due to a desensitization at all levels of the HPA axis (Sapolsky and Meaney, 1986). The suppression of the pups’ stress reactivity is tightly controlled by maternal care and can only be disrupted by severe stressors, such as removal of the dam (Levine, 2002; Schmidt et al., 2003). The SHRP coincides with a critical period of postnatal brain development, and its evolutionary purpose is most likely to minimize the damaging effects of glucocorticoids on the developing brain (Sapolsky and Meaney, 1986). In HR mice, the fragmented maternal care induced by the ELS paradigm was apparently sufficient to disrupt the suppression of the stress response system, leading to elevated basal CORT levels during the SHRP. As the GR-mediated negative feedback loop is not yet functioning in pups at this young age (Meaney et al., 1985), the CORT levels remained elevated with the potential to interfere with neuroendocrine receptor expression and to activate epigenetic processes in the developing brain. Once the negative feedback loop became functionally instantiated, basal CORT levels of HR ELS pups dropped to normal levels, while lasting epigenetic programming effects on the neuroendocrine system may have led to the augmented stress reactivity measured in adult HR ELS animals.

In contrast, adult LR mice showed no differences in their stress response associated with ELS rearing conditions, again replicating earlier results (McIlwrick et al., 2016). However, we noted that the baseline CORT levels of LR ELS mice increased significantly from early to late adulthood (pairwise comparisons LR ELS early vs. late adulthood: p = 0.031), and overall, LR mice displayed a rise in their stress response CORT levels in late adulthood (pairwise comparisons LR early vs. late adulthood: p = 0.051). To date, only some acute (pre-weaning) and no lasting effects of ELS exposure have been reported in LR animals (McIlwrick et al., 2016). Our new data now suggests that there are indeed some lasting consequences, but that these only appear with a late adulthood onset. Strikingly, the ELS-induced effects in LR mice, while not being very pronounced, seem to be rather favorable in nature: aged LR ELS animals showed signs of improved cognitive function and had slightly raised baseline CORT levels compared to LR STD mice. Since the effects of stress hormones have an inverted U-shaped relationship to cognition (de Kloet et al., 1999; Mateo, 2008; Sapolsky, 2015) and LR mice usually have a very low baseline HPA axis tone (Touma et al., 2008, 2009), a small increase in baseline activation may convey some beneficial aspects for attention and behavioral reactivity, by increasing the relative occupancy of MRs compared to GRs in the hippocampus (de Kloet et al., 1999; Herbert et al., 2006; Ferguson and Sapolsky, 2007) and thus promote cognitive function.

In early adulthood, Bdnf levels were downregulated in the dorsal and ventral hippocampus of HR ELS mice compared to HR STD animals (Figures 7A,E), likely reflecting a downstream effect of exaggerated glucocorticoid stimulation, as it has repeatedly been shown that the BDNF- and the glucocorticoid-signaling pathways are closely interlinked and show bi-directional cross talk (Jeanneteau and Chao, 2013; Daskalakis et al., 2015). In late adulthood, the difference in Bdnf levels between HR ELS and STD mice was reduced, due to a downward shift in the expression levels of HR STD animals (Figures 7B,F). Overall, HR mice had lower Bdnf levels than LR mice, confirming previous findings at the level of proteins in the hippocampus in HR animals (Knapman et al., 2010b). Being an important mediator of neural growth and survival, as well as of synaptic plasticity (Huang and Reichardt, 2001; Daskalakis et al., 2015), BDNF plays a central role in the underlying processes of learning and memory (Cunha et al., 2010). Thus, a reduced availability of BDNF, as observed in HR mice and in HR ELS in particular, may contribute to the impaired cognitive performance of these animals. ELS is known to impact on BDNF expression in the hippocampus (Liu et al., 2000), reinforcing the association with ELS-induced cognitive deficits that have been described (Daskalakis et al., 2015). In line with this, our data showed a significant positive correlation of BDNF mRNA expression in the dorsal hippocampus and cognitive performance in the object recognition test. The reason this association was not significant in the Y-maze test may be that spatial discrimination in the Y-maze test relies mainly on the hippocampal formation, while the object recognition test performance also depends on the prefrontal and parahippocampal cortex, where BDNF is a key regulator of neuronal function.

The reduced hippocampal expression of BDNF mRNA most likely reflects a dynamic epigenetic signature of ELS and augmented stress hormone signaling, rather than a genetically encoded difference between the mouse lines, since in early adulthood, HR ELS and STD mice showed condition-dependent differences in Bdnf expression levels. Glucocorticoids can have profound effects on the regulation of a range of transcription factors, the epigenome, and mircoRNAs, suggesting a wide array of potential programming pathways (de Kloet et al., 2009; McGowan et al., 2009; Suri and Vaidya, 2013).

Changes in the CRH system have also repeatedly been implicated in the adverse effects of ELS (Avishai-Eliner et al., 2001; Rice et al., 2008; Ivy et al., 2010; Korosi et al., 2010; Wang et al., 2013; Fuge et al., 2014; Liao et al., 2014). In early adulthood, there was an upregulation of Crh in the dorsal hippocampus of ELS-exposed HR mice compared to HR STD animals (McIlwrick et al., 2016, also shown in Figure 8A), which we now observed to be stable into late adulthood (Figure 8B). However, at both ages, we detected no evidence for changes in the expression of Crhr1 in the hippocampus (McIlwrick et al., 2016, and Figures 8C,D,G,H).

An increased tone of CRH activity in the hippocampus could constitute a further pathway contributing to deficits in hippocampus-dependent cognitive tasks, as CRH-signaling has been implicated in dendritic remodeling (Chen et al., 2008, 2012) and is a prominent target for ELS-induced epigenetic programming (Murgatroyd and Spengler, 2011; Karsten and Baram, 2013). The significant negative correlation of dorsal hippocampal Crh expression and cognitive performance in aged animals provides evidence for this association. As discussed above, task performance in the Y-maze test is highly dependent in intact hippocampal functioning, while the object recognition test also recruits different cortical areas. This explains the stronger impact of increased hippocampal Crh expression on the spatial memory in the Y-maze than on recognition memory in the object recognition test. In LR mice, no ELS-associated changes in Crh and Crhr1 expression were observed in the dorsal hippocampus during early adulthood (McIlwrick et al., 2016, also shown in Figures 8A,C), but a downregulation of both genes became evident in late adulthood (Figures 8B,D), coinciding with somewhat enhanced cognitive performance of LR ELS mice (Figure 4B). In conclusion, our data suggests that ELS exposure triggers a range of neuroendocrine and molecular alterations, the effects of which emerge gradually in early adulthood and strongly depend on the animal’s inherited predisposition for high or low stress reactivity.

The selection of appropriate behavioral tests is critical to reliably measure small effects of experimental manipulations in behaving animals. To assess cognitive function in early and late adulthood we used the Y-maze test at both time points, but supplemented this with different behavioral tasks (i.e., the WCM was used only in early adulthood; the object recognition test was used only in late adulthood), which creates some asymmetry in the data. The reason why we did not repeat the WCM in the late adulthood animal cohort was that this test presupposes the animal’s ability to navigate using visual cues and albino mice are poorly equipped for vision-based tasks (Brown and Wong, 2007). Hence, animals of the SR mouse model in general performed relatively poorly compared to, e.g., wild type C57Bl6/N mice (Kleinknecht et al., 2012) in the WCM. In addition, the WCM is a relatively stressful test, due to the need for the animals to swim in water, which may impact on the animal’s performance. Therefore, we decided to use a less stressful and less vision-dependent test in the late adulthood cohort of animals. Since the Y-maze test was identical at both time points and the results concurred well with the results of both other cognitive tasks, we believe that our conclusions regarding changes in cognitive function from early to late adulthood are nonetheless valid and justifiable.

In the presented work, we investigated the effects of ELS in male mice, only. However, the clinical reality shows that women are at a twofold increased risk for affective and stress-related disorders (Gater et al., 1998; Castle, 2007). Several factors may play a role in this enhanced vulnerability, including differences in neuroendocrine regulation and interaction of reproductive hormones (Seeman, 1997; Young et al., 2001; Bale, 2006), which may be exacerbated through ELS experiences. As the findings from male animals are not necessarily directly transferable to females and therefore only provide limited information regarding large parts of the patient population, future studies should include female subjects in the investigation.

In recent years, several gene variants that contribute to individual risk or resilience have been identified, including Nr3c1 (Wust et al., 2009), Nr3c2 (DeRijk et al., 2006), Fkbp5 (Ising et al., 2008), Crhr1 (Clarke and Schumann, 2009), Crhbp (Wang et al., 2007), Gabra6 (Uhart et al., 2004), and Slc6a4 (Way and Taylor, 2010). It would contribute to our understanding of the gene × environment interaction described here to have a better knowledge about the genomic sequence of the three SR mouse lines, in order to seek confirmation for some of the known risk polymorphisms and to detect new potential candidates. Moreover, an analysis of the methylation status of candidate genes may add valuable information about epigenetic changes induced by ELS in the three SR mouse lines.