The effects of brain serotonin deficiency on behavioural disinhibition and anxiety-like behaviour following mild early life stress

Abstract

Aberrant serotonin (5-HT) signalling and exposure to early life stress have both been suggested to play a role in anxiety- and impulsivity-related behaviours. However, whether congenital 5-HT deficiency × early life stress interactions influence the development of anxiety- or impulsivity-like behaviour has not been established. Here, we examined the effects of early life maternal separation (MS) stress on anxiety-like behaviour and behavioural disinhibition, a type of impulsivity-like behaviour, in wild-type (WT) and tryptophan hydroxylase 2 (Tph2) knock-in (Tph2KI) mice, which exhibit ∼60–80% reductions in the levels of brain 5-HT due to a R439H mutation in Tph2. We also investigated the effects of 5-HT deficiency and early life stress on adult hippocampal neurogenesis, plasma corticosterone levels and several signal transduction pathways in the amygdala. We demonstrate that MS slightly increases anxiety-like behaviour in WT mice and induces behavioural disinhibition in Tph2KI animals. We also demonstrate that MS leads to a slight decrease in cell proliferation within the hippocampus and potentiates corticosterone responses to acute stress, but these effects are not affected by brain 5-HT deficiency. However, we show that 5-HT deficiency leads to significant alterations in SGK-1 and GSK3β signalling and NMDA receptor expression in the amygdala in response to MS. Together, these findings support a potential role for 5-HT-dependent signalling in the amygdala in regulating the long-term effects of early life stress on anxiety-like behaviour and behavioural disinhibition.

Animal models, anxiety, behavioural science, early life stress, impulsivity, serotonin, tryptophan hydroxylase

Introduction

Early life stress has been proposed to be a key factor in the development of psychiatric illness (Heim and Nemeroff, 2001) and thus the development of animal models of early life stress has become an area of active research. One of the most common rodent models of early life stress is the repeated maternal separation (MS) stress model, which has been reported to induce anxiety-like behaviour in rodents (Huot et al., 2001; Romeo et al., 2003) and non-human primates (Young et al., 1973). However, despite the initial optimism regarding the use of MS to model anxiety-like behaviour, recent work has failed to replicate the anxiety-producing effects of MS in several strains of inbred mice (Millstein and Holmes, 2007; Parfitt et al., 2007), suggesting that MS may represent a relatively mild form of stress that does not elicit robust anxiety-like behaviour. Nonetheless, we hypothesized that MS might be an appropriate model with which to screen for genetic factors that increase stress vulnerability.

Previous research has implicated serotonin (5-HT) in the regulation of stress sensitivity (Rioja et al., 2004; van den Hove et al., 2011) and has revealed a role for the brain 5-HT system in aberrant emotional behaviour. For example, mutations in 5-HT system genes have been identified in individuals with anxiety disorders (Lesch et al., 1996) and severe impulsivity (Bevilacqua et al., 2010). Also, reduced levels of the 5-HT metabolite, 5-hydroxyindoleacetic acid (5-HIAA), have been reported in the cerebrospinal fluid of patients with impulsive-aggressive personality disorders, suggesting that they may exhibit 5-HT deficiency (Kruesi et al., 1990). One 5-HT-system gene that has been implicated in anxiety and impulsive behaviour is tryptophan hydroxylase-2 (Tph2; Kim et al., 2009; Perez-Rodriguez et al., 2010), the rate-limiting enzyme for brain 5-HT synthesis. The R441H polymorphism in Tph2, which has been shown to retard 5-HT synthesis both in vitro and in vivo (Zhang et al., 2005; Beaulieu et al., 2008; Jacobsen et al., 2012), was first identified in a cohort of elderly depression patients (Zhang et al., 2005). Mice harbouring the analogous mutation (R439H) in murine Tph2 [Tph2 knock-in (TphKI) mice] have reduced levels of brain 5-HT and display increased aggression-, depression- and anxiety-like behaviours (Beaulieu et al., 2008; Jacobsen et al., 2012).

The increased impulsive aggression-like phenotype observed in Tph2KI animals is consistent with the phenotypes reported in several other genetic models of 5-HT dysfunction (Brunner and Hen, 1997; Angoa-Perez et al., 2012; Mosienko et al., 2012). Interestingly, several previous studies have indicated that the impulsivity-like phenotypes in animal models of serotonergic dysfunction are motor in nature, not cognitive (Brunner and Hen, 1997; Angoa-Perez et al., 2012). Thus, for the present study, we have focused on the effects of early life stress on motor impulsivity-like behaviours and behavioural inhibition in 5-HT-deficient mice. To begin to investigate the potential cellular and molecular mechanisms underlying any observed behavioural differences, we also examined hippocampal neurogenesis and signal transduction in the amygdala (Amyg), focusing on several pathways that have been implicated in anxiety-likebehaviour and/or stress responsiveness, including the GSK3β, ERK1/2 and SGK-1 pathways, and expression of the N-methyl-d-aspartate (NMDA) receptor subunit, NR1 (Roceri et al., 2002; Beaulieu et al., 2008; Kaji and Nonogaki, 2010; Mines et al., 2010; Ishikawa et al., 2012; Kinoshita et al., 2012).

Method

Animals

The generation and breeding of Tph2 R439H KI mice, which are on a mixed background (c57BL6/J–129S6/SvEv), has been described previously (Beaulieu et al., 2008). After weaning, mice were housed two to five per cage, maintained at 23±2°C on a 12 h light–dark cycle (lights on 07:00 hours) and provided food and water ad libitum. All experiments were conducted with an approved protocol from the Duke University Institutional Animal Care and Use Committee.

Maternal separation

Wild-type (WT) and Tph2KI pups were separated from their dams for 3 h each day on post-natal days 1–14. During this time, the pups were placed on a heating pad and remained in contact with their littermates. Control pups were handled briefly for 5 min and returned to the nest with their littermates and dam. The first cohort used in this study consisted of WT and homozygous mutant Tph2KI mice as dams, but all remaining cohorts used heterozygous Tph2KI mice as dams to control for potential genotype differences in maternal care. No statistically significant differences among cohorts were observed and data from all cohorts were combined. Litters were not culled, and both male and female animals were used for subsequent behavioural analyses starting at age 8 wk. No significant sex differences were observed, and data from both sexes were combined. Age-matched WT and Tph2KI littermates were used for all experiments. Mice were examined in multiple paradigms, with at least a 1-wk interval between tests. The time-line of behavioural testing for each cohort of mice is shown in Supplementary Table 1.

Mouse behavioural tests

The novel open field (NOF), elevated zero maze (EZM) and novelty suppressed feeding (NSF) tests were conducted as described previously (Fukui et al., 2007). For the NOF, mice were placed in a NOF (21 × 21 × 30 cm) and the total distance travelled, the distance travelled in the centre of the arena and the time spent in the centre of the arena were measured in 5 min intervals over a 30 min period using VersaMax software.

For the EZM, mice were placed into a closed arm of the maze and were videotaped using a camera suspended above the maze. Each mouse was allowed to freely investigate the apparatus for 5 min. The amount of time each mouse spent in the open and closed arms of the maze, the numbers of open arm entries and the number of stretch-attend postures were subsequently determined using the Observer program (Noldus, USA).

For the NSF, animals were food deprived for 24 h prior to testing but continued to have ad libitum access to water. A single food pellet was placed in the centre of a 60 × 40 cm brightly lit open field (∼1300 lux). Mice were placed in the corner of the open field and the latency to begin consuming the pellet was recorded. Mice that did not consume the food pellet within 5 min were not included in the analysis.

Cliff-avoidance test

The cliff-avoidance test was conducted essentially as described previously (Matsuoka et al., 2005). Briefly, mice were placed on a transparent glass platform (12 cm in diameter, at an elevation of 20 cm). The latency to jump off the platform head-first onto a pad below and the number of head-dips (for mice that remained on the platform for at least 100 s) were recorded. Mice that fell backwards from the platform were not included in the analysis.

Step-down passive avoidance test

The step-down passive avoidance test was conducted as described previously (Prado et al., 2006), but only short-term avoidance was tested, immediately following conditioning.

The latency for each mouse to step down from the platform and the number of grid touches each mouse made with its forepaws prior to stepping down from the platform were recorded.

Fear conditioning

Fear conditioning was performed as described previously (Schmalzigaug et al., 2009). Briefly, fear conditioning was conducted in a mouse fear conditioning chamber (Med Associates, USA). Mice were initially allowed to freely explore the apparatus for 2 min, after which they were presented with the conditioned stimulus (a 72 dB, 29 kHz tone) for 30 s. Each mouse received a 0.4 mA foot shock for 2 s just prior to tone termination. Each mouse was allowed to remain in the chamber for an additional 30 s and was then returned to its home cage. After 24 h, each mouse was individually returned to the conditioning chamber, and its freezing behaviour was recorded over a 5 min period in the absence of the tone or foot-shock.

Evaluation of plasma corticosterone

Plasma corticosterone levels were determined using a competitive binding enzyme immunoassay kit according to the manufacturer's instructions (Assay Designs, USA). The sensitivity of the assay was 26.99 pg/ml, and the intra-assay variability was 6.6–8.4%. For acute restraint-stress studies, mice were restrained within a ventilated 50 ml Falcon tube for 15 min at approximately 14:00 hours. Immediately following restraint, animals were decapitated and trunk blood was collected, placed on ice, and centrifuged for 10 min. Control mice were killed without being subjected to restraint. Plasma was collected and examined for corticosterone levels.

Western blotting

The brains of mice killed for the determination of plasma corticosterone were collected for Western blot analysis. Following acute restraint (described earlier), 2 mm punches from the nucleus accumbens (NAc), hippocampus (Hip, primarily dentate gyrus), and Amyg were collected and snap-frozen in liquid nitrogen. Western blots were performed according to standard protocols (Beaulieu et al., 2008). Bilateral punches from the same animal were pooled and considered a single sample (i.e. n = 1). Quantification was performed using densitometry with the Image J program (National Institutes of Health). The relative levels of proteins were normalized against GAPDH (or total, non-phospho-specific ERK or GSK3β) as a loading control.

Antibodies used were: rabbit anti-NR1 (sc-9058, Santa Cruz; 1:500); mouse anti-GAPDH (MAB374, Millipore; 1:1000); rabbit anti-phospho-GSK3β (#9323, Ser⁹, Cell Signalling; 1:500); mouse anti-total GSK3α/β (sc-7291, Santa Cruz, 1:500); rabbit anti-phospho-p44/p42 MAPK (Erk1/2; #9101, Tyr^202/204, Cell Signalling; 1:500); mouse anti-total ERK (#9107, Cell Signalling, 1:500); rabbit anti-phospho-SGK1 (#441260G, Tyr²⁵⁶, Invitrogen; 1:500). The appropriate AlexaFluor 680 or AlexaFluor 800- conjugated secondary antibodies were used (Molecular Probes; 1:10,000 dilution), and blots were developed using an Odyssey LI-COR system (LI-COR Biosciences, USA).

Immunohistochemistry

Mice were injected with 5-bromo-2′-deoxyuridine (BrdU) three times (100 mg/kg i.p., once each 24, 16 and 4 h prior to euthanasia) and perfused with 10% neutral buffered formalin (NBF). Brains were removed, post-fixed in NBF overnight, cryopreserved in 30% sucrose in PBS for 48 h, embedded in optimal cutting temperature compound (VWR International, USA), and cut into 20 µm sections using a cryostat. Alternate sections throughout the Hip were collected and mounted onto glass slides. Every tenth section collected was processed for immunohistochemistry (IHC) and both hemispheres were examined.

For BrdU stainings, sections were dehydrated in two consecutive 5 min washes in xylene substitute (Sigma, USA) then rehydrated through a series of washes in ethanol (EtOH): 100% EtOH for 10 min; 95% EtOH for 3 min; 85% EtOH for 3 min; 70% EtOH for 3 min. Slides were washed in PBS for 5 min and subjected to sodium citrate antigen retrieval. Briefly, sections were submerged in sodium citrate antigen retrieval buffer (10 mm sodium citrate and 10 mm citric acid, pH 6.0) and boiled in a microwave for 20 min. Sections were placed at 4°C for 20 min and washed in PBS for 5 min. Sections were then incubated for 1 h in 2 n hydrochloric acid in PBS and washed three times in PBS (5 min each). For doublecortin (DCX) staining, these antigen retrieval steps were omitted, and sections were thawed and rinsed once in PBS for 5 min. All sections were then blocked in 5% bovine serum albumin (BSA) in PBS containing 0.1% Triton (PBS-t) for 1 h. After three washes in PBS, sections were incubated overnight in primary antibody diluted in 1% BSA in PBS-t. Rat anti-BrdU (OBT0030G, Accurate Chemical Corporation, USA, 1:200) and rabbit anti-DCX (ab18723, Cell Signaling, 1:200) primary antibodies were used. Sections were washed three times in PBS and then incubated for 1 h at room temperature with Alexafluor 568-conjugated anti-rat or Alexafluor 488-conjugated anti-rabbit secondary antibodies (Life Technologies, USA) diluted 1:500 in 5% BSA in PBS-t. Sections were then washed three times in PBS and coverslipped using SlowFade Gold Antifade Reagent with DAPI (Life Technologies, USA). Non-overlapping images of the entire subgranular zone were taken on a fluorescence microscope (Zeiss) by an individual blinded to the genotype and history of stress exposure of the animals. The number of BrdU+ and DCX+ cells were counted by a blinded observer. One set of brain sections corresponding to every tenth section that was collected (five sections per mouse) was processed for BrdU IHC, and a second set of sections was processed for DCX IHC. Both hemispheres were examined, and the number of BrdU+ or DCX+ cells was then multiplied by 10 to obtain the value for each mouse.

For confocal analysis, sections were stained according to the above protocol, except that they were counterstained with TOTO-3, a nuclear marker, for 15 min (1:500 dilution, Life Technologies, USA) prior to coverslipping with SlowFade Gold Antifade Reagent without DAPI (Life Technologies, USA).

Measurements of 5-HT tissue content

The levels of 5-HT and 5-HIAA in the Amyg were quantified essentially as described previously for other brain regions (Jacobsen et al., 2012). In brief, mice were rapidly decapitated; Amyg tissues were collected (see earlier) and stored at −80°C. Tissues were thawed on ice, homogenized by sonication in approximately 10 volumes of ice cold 0.1 m HClO₄. One aliquot of the homogenate was used for protein determination using Pierce BCA assay kits according to the manufacturer's protocol. A second aliquot of the homogenate was diluted five times more in 0.1 m HClO₄, vortexed, and debris and membranes pelleted by centrifugation (20000 g). The supernatant was recovered and filtered through 0.1 µm centrifugal filters (Ultrafree, EMD Millipore, USA). High pressure liquid chromatography with electrochemical detection was used to quantify 5-HT and 5-HIAA levels in the supernatant as described previously (Jacobsen et al., 2012). Levels of 5-HT and 5-HIAA were expressed as ng/mg protein.

Statistical analysis

Statistical analyses were performed using JMP software (SAS, USA) using two- and three-way analysis of variance (ANOVA) with Tukey's or Bonferroni's post hoc tests for multiple comparisons or Pearson's χ², where appropriate. To compare the levels of 5-HT and 5-HIAA in WT and Tph2KI mice t tests were applied, using JMP software.

Results

MS differentially affects anxiety- and impulsivity-like behaviours in WT and Tph2KI mice

The NOF test is commonly used to measure locomotor activity and anxiety-like behaviour (Treit and Fundytus, 1988). In this test, no significant main or interaction effects of genotype or MS were observed on overall locomotor activity (Fig. 1a). However, an examination of the distance travelled in the centre of the open field revealed a significant genotype × early life stress interaction (F_1,41 = 7.483, p = 0.009; Fig. 1b). Under control conditions, Tph2KI mice were less active in the centre zone than WT controls (p = 0.006), a finding that is consistent with Tph2KI mice exhibiting a baseline anxiety-like phenotype (Beaulieu et al., 2008). MS led to a decrease in centre activity (i.e. increased anxiety-like behaviour) in WT animals compared to non-stressed WT mice (p = 0.023). When the amount of time spent in the centre zone was analysed, Tph2KI animals were observed to spend less time in the centre zone than their WT littermates (main effect of genotype: F_1,41 = 6.3217, p = 0.0159; Fig. 1c). Interestingly, a significant genotype × stress interaction was also observed (F_1,41 = 10.435, p = 0.0024), and post hoc tests revealed that control Tph2KI mice spent less time in the centre zone than WT controls (p < 0.001), whereas no genotype differences were observed in MS-exposed animals. MS reduced centre time in WT mice (p = 0.01), but MS-exposed Tph2KI animals actually exhibited a trend toward increased centre time compared to control Tph2KI mice (p = 0.055). These results demonstrate that MS leads to significantly different effects on anxiety-like behaviour in WT and Tph2KI mice.

Fig. 1

Maternal separation (MS) stress leads to differential effects on anxiety-like behaviour in wild-type (WT) and tryptophan hydroxylase 2 knock-in (Tph2KI) mice. (a) MS has no significant effects on the total distance travelled in 30 min in the novel open field in control (Con) or MS WT and Tph2KI mice. (b) MS significantly decreases distance travelled in the centre in WT but not Tph2KI mice. (c) MS significantly decreases time spent in the centre zone in WT but not Tph2KI mice. (d) The percentage time spent in the open areas of the elevated zero maze is differentially affected by MS in WT and Tph2KI mice. Similarly, the number of open arm entries (e) and the number of stretch-attend postures (f) are differentially affected by stress in WT and Tph2KI mice during the 5 min zero maze test. For (a–c): WT Con, eight males and five females; Tph2KI Con, four males and five females; WT MS, five males and eight females; Tph2KI MS, five males and four females. For (d–f): WT Con, 14 males and eight females; Tph2KI Con, eight males and eight females; WT MS, nine males and 13 females; Tph2KI MS, nine males and 10 females. * p < 0.05 for a within-genotype effect of MS; # p < 0.05 for a between-genotype effect within stress condition.

To further explore the effects of 5-HT deficiency on anxiety-like responses to MS, we examined mice in the EZM, a test in which the percentage time in the open areas is an inverse measure of anxiety-like behaviour (Shepherd et al., 1994). Two-way ANOVAs revealed significant genotype × early life stress interactions for the percentage time in the open areas (F_1,76 = 4.6705, p = 0.0338; Fig. 1d), the number of open arm entries (F_1,76 = 8.3360, p = 0.0051; Fig. 1e) and the number of stretch-attend postures (F_1,76 = 5.2692, p = 0.0246; Fig. 1f). However, the magnitude of these effects was relatively small, and Tukey's post hoc tests did not reveal any significant differences between individual groups on any of these parameters. Taken together, these results indicate that MS leads to relatively mild effects on anxiety-like behaviour and confirm that 5-HT deficiency can modify anxiety-like behavioural responses to early life stress.

Because 5-HT has been implicated in the regulation of impulsivity, we next evaluated the effects of MS on behavioural inhibition in WT and Tph2KI mice. In the cliff-avoidance test, which has been used to measure impulsivity-like behaviour (Matsuoka et al., 2005), Tph2KI mice exposed to MS exhibited significantly more jumping from the elevated platform than any other group (χ² = 13.688, n = 64, d.f. = 3, p = 0.0003; Fig. 2a). Specifically, eight of the 18 MS Tph2KI mice jumped off the platform within the 5-min test, compared to none of the 14 WT controls, one of the 12 Tph2KI controls, and two of the 20 MS-exposed WT mice. Thus, MS-exposed Tph2KI animals appeared to exhibit behavioural disinhibition, a subtype of impulsivity-like behaviour. When exploratory head-dips were examined, a significant main effect of genotype (F_1,26 = 8.9519, p = 0.006) and a significant genotype × early life stress interaction (F_1,26 = 4.5242, p = 0.0431) were observed. Specifically, MS increased the number of head-dips in Tph2KI mice compared to control Tph2KI animals (p < 0.05), but no significant effects were observed in WT animals. Consequently, MS-exposed Tph2KI mice exhibited increased numbers of head-dips compared to MS-exposed WT animals (p < 0.05, Fig. 2b).

Fig. 2

Maternal separation (MS) stress increases behavioural disinhibition in tryptophan hydroxylase 2 knock-in (Tph2KI) mice. (a) A cumulative jumping-response curve shows that Tph2KI mice subjected to MS exhibit more jumping events and shorter jumping latencies than control (Con) Tph2KI mice and Con- and MS wild-type (WT) mice in the cliff-avoidance test. (b) Tph2KI mice subjected to MS also exhibit increased numbers of head-dips in the cliff-avoidance test compared to the other groups. (c) MS also led to differential effects on the latency to step down in the step-down passive avoidance test in WT and Tph2KI mice. (d) Tph2KI mice exhibited a marked reduction in the number of grid touches prior to stepping down from the elevated platform in the step-down passive avoidance test regardless of stress exposure. (e) The percentage time spent immobile during contextual fear conditioning was significantly reduced by MS in both genotypes. (f) The feeding latencies of control and MS WT and Tph2KI mice in the novelty suppressed feeding test were differentially affected by MS, with MS Tph2KI mice exhibiting significantly reduced latency compared to the other groups. For (a): WT Con, six males and eight females; Tph2KI Con, five males and seven females; WT MS, 11 males and nine females; Tph2KI MS, 10 males and eight females. For (b): WT Con, three males and two females; Tph2KI Con, three males and three females; WT MS, seven males and four females; Tph2KI MS, five males and three females. For (c–e): WT Con, four males and four females; Tph2KI Con, four males and five females; WT MS, 10 males and five females; Tph2KI MS, six males and four females. For (f): WT Con, four males and four females; Tph2KI Con, four males and five females; WT MS, 11 males and five females; Tph2KI MS, six males and four females. * p < 0.05 for a within-genotype effect of MS; # indicates p < 0.05 for a between-genotype effect within stress condition; ** p = 0.003 by Pearson's χ² analysis, χ² = 13.688; ^ significant main effect of MS, p = 0.0025 by two-way analysis of variance.

In the step-down passive avoidance test, which can be used as a measure of behavioural disinhibition or impulsivity-like behaviour (Matzel et al., 2008), Tph2KI animals exhibited an overall reduction in step-down latencies compared to WT mice (main effect of genotype; F_1,39 = 5.9840, p = 0.019; Fig. 2c). A significant genotype × early life stress interaction was also observed (F_1,39 = 4.6956, p = 0.0364), with control Tph2KI mice exhibiting a shorter latency to step down than WT controls (p = 0.0234, Fig. 2c). In contrast, no significant genotype differences in step-down latency were observed after MS. When precautionary behaviour was examined, WT mice used their front paws to make contact with the grid prior to stepping down from the elevated platform much more frequently than Tph2KI mice (main effect of genotype: F_1,39 = 9.5773, p = 0.0036), but MS had no effect on this behaviour (Fig. 2d). Together, these results indicate that brain 5-HT deficiency can lead to behavioural disinhibition. To confirm that the dramatic decrease in step-down latency observed in control Tph2KI mice was not due to a memory deficit or insensitivity to foot-shocks, we conducted fear conditioning in control and MS-exposed WT and Tph2KI mice. Control WT and Tph2KI mice exhibited similar freezing responses in contextual fear conditioning, indicating that both genotypes can learn to respond to noxious stimuli (Fig. 2e). Interestingly, MS led to reduced freezing behaviour after contextual fear conditioning (main effect of MS: F_1,40 = 10.3681, p = 0.0025; Fig. 2e), thus suggesting that MS may lead to impaired contextual fear learning and memory.

The NSF test has also been used as an index of behavioural disinhibition or impulsivity-like behaviour (Bevilacqua et al., 2010; Angoa-Perez et al., 2012) and reduced feeding latency in this test has been associated with behavioural disinhibition. A two-way ANOVA revealed a significant main effect of genotype (F_1,87 = 6.4589, p = 0.0128) and a significant genotype × stress interaction (F_1,87 = 4.0895, p = 0.0462; Fig. 2f). Tph2KI mice subjected to MS showed reduced feeding latencies compared to MS-exposed WT animals (p = 0.0061). Taken together, these results indicate that MS induces significant behavioural disinhibition in Tph2KI, but not in WT animals.

Endocrine and neurogenic responses to MS

Previous research suggests that MS potentiates acute stress-induced increases in plasma corticosterone in adult rodents (Murgatroyd et al., 2009). To determine whether 5-HT deficiency regulates this effect of MS, we measured the resting (afternoon) and acute restraint-induced levels of plasma corticosterone in control and MS WT and Tph2KI mice. A three-way ANOVA revealed a significant main effect of acute restraint (F_1,37 = 157.5649, p < 0.0001) and a significant early life stress × restraint interaction (F_7,37 = 7.3781, p = 0.01; Fig. 3), with acute restraint leading to higher corticosterone levels in MS animals than in controls (p = 0.014). MS did not alter the resting plasma levels of corticosterone in either genotype and no significant genotype differences were observed under any conditions (Fig. 3).

Fig. 3

Maternal separation (MS) stress potentiates the hypothalamic–pituitary–adrenal axis reactivity in response to acute stress. Baseline and acute restraint stress-induced levels of plasma corticosterone in control (Con) and MS (wild-type) WT and tryptophan hydroxylase 2 knock-in (Tph2KI) mice. A significant interaction between MS and restraint and a main effect of restraint were observed. WT Con, four males and four females; Tph2KI Con, two males and three females; WT MS, three males and four females; Tph2KI MS, two males and one female; WT restraint, two males and three females; Tph2KI restraint, two males and three females; WT MS restraint, three males and two females; Tph2KI MS restraint, two males and two females. * p <0.05 for the Con vs. the MS restraint stress comparison by two-way analysis of variance.

In addition to potentiating corticosterone responses, early life stress has also been shown to inhibit adult hippocampal neurogenesis (Mirescu et al., 2004; Kikusui et al., 2009). We observed no statistically significant baseline genotype differences in BrdU incorporation in unstressed animals, but MS led to a slight, but significant, overall decrease in BrdU incorporation (main effect of stress: F_1,58 = 4.8511, p = 0.0316; Fig. 4). Although this effect appeared to be more pronounced in WT animals, the interaction was not significant. IHC for DCX did not reveal any statistically significant effects of MS on the number of immature neurons in either genotype (Supplementary Fig. S1), suggesting that the effects of MS on neurogenesis in mice are relatively minor and unaffected by brain 5-HT deficiency. Thus, it does not appear that differential corticosterone or neurogenic responses to MS underlie the observed behavioural differences following MS in WT and Tph2KI animals.

Fig. 4

Maternal separation (MS) stress decreases adult hippocampal neurogenesis in wild-type (WT) and serotonin-deficient animals. (a) Quantification of the number of 5-bromo-2′-deoxyuridine (BrdU)+ cells in the subgranular layer of the dentate gyrus in control (Con) and MS WT and tryptophan hydroxylase 2 knock-in (Tph2KI) mice. Representative images from WT Con (b), Tph2KI Con (c), WT MS (d) and Tph2KI MS (e) are shown. BrdU is shown in red, and the nuclear marker TOTO-3 is shown in blue. WT Con, 10 males and six females; Tph2KI Con, nine males and seven females; WT MS, six males and nine females; Tph2KI MS, eight males and seven females. The scale bar indicates 20 µm. A significant main effect of early life stress on BrdU incorporation (* p < 0.05 by two-way analysis of variance) was observed.

Signalling responses to MS

To determine whether 5-HT deficiency or MS would affect signal transduction in the brain in response to acute stress in adulthood, we examined several signalling pathways that have been reported to play a role in stress responses, including the SGK-1 (Kaji and Nonogaki, 2010), ERK1/2 (Ishikawa et al., 2012) and GSK3 (Kinoshita et al., 2012) pathways, the last of which has also been previously implicated in the behavioural effects of 5-HT deficiency (Beaulieu et al., 2008). An examination of the effects of MS on phosphorylated (p) GSK3β in the Amyg revealed a significant genotype × stress interaction (F_1,16 = 6.5426, p = 0.0211), with MS-exposed Tph2KI mice exhibiting reduced levels of pGSK3β compared to MS-exposed WT animals (p = 0.047, Fig 5a). No significant effects of genotype or MS on pGSK3β were observed in the Hip or the NAc (B. D. Sachs, unpublished observations). In contrast, no significant differences were observed in ERK1/2 phosphorylation in the Amyg as a consequence of MS or 5-HT deficiency (Fig. 5b). Regarding SGK-1, Tph2KI mice exhibited an overall reduction in the levels of phosphorylated SGK-1 in the Amyg when compared to WT controls (main effect of genotype: F_1,16 = 16.6569, p = 0.0009; Fig. 5c).

Fig. 5

Signal transduction effects of maternal separation (MS) stress in wild-type (WT) and tryptophan hydroxylase 2 knock-in (Tph2KI) mice. Quantification of Western blots for phosphorylated (p) GSK3β (a), pERK (b), pSGK-1 (c) and NR1 (d), in the amygdala. Significant interactions between MS and genotype were observed in (a) and (d). Significant main effects of genotype were observed for pSGK-1 and NR1. * Indicates p < 0.05 for a significant main effect of genotype by two-way analysis of variance; # Indicates p < 0.05 for a between-genotype effect within stress condition. n = 5 per group (three males and two females in each).

No significant differences in pSGK-1 were observed in the Hip (B. D. Sachs, unpublished observations), but MS induced a significant reduction in the levels of pSGK-1 in the NAc in both genotypes (main effect of MS: F_1,16=p = 0.0033; Supplementary Fig. S2).

Early life stress is known to alter the expression levels of NMDA receptors in rat brain (Roceri et al., 2002). In our study, we observed a significant reduction in the levels of the NR1 subunit in the Amyg of Tph2KI mice when compared to WT animals (main effect of genotype: F_1,16 = 12.2606, p = 0.003; Fig. 5d). In addition, we also observed a significant genotype × early life stress interaction (F_1,16 = 4.8191, p = 0.0432), with MS-exposed Tph2KI mice exhibiting significantly less NR1 expression than MS-exposed WT animals (p = 0.0048, Fig. 5d). In contrast to the Amyg, no significant group differences in NR1 expression were observed in the Hip or NAc (B. D. Sachs, unpublished observations).

Neurochemical effects of the Tph2 R439H mutation in the Amyg

To confirm that Tph2KI mice exhibit 5-HT deficiency in the Amyg, as has been previously reported for the frontal cortex, Hip and striatum (Beaulieu et al., 2008), we compared the tissue levels of 5-HT and 5-HIAA in WT and Tph2KI mice. An examination of the tissue content of 5-HT (t = 8.636, p < 0.0001, n = 8 per group) and 5-HIAA (t = 6.777, p < 0.0002, n = 8 per group) in the Amyg revealed that Tph2KI mice exhibit a 65–70% reduction in 5-HT in this brain area as well (Fig. 6).

Fig. 6

Levels of serotonin (5-HT) and 5-hydroxyindoleacetic acid (5-HIAA) in the amygdala of wild-type (WT) and tryptophan hydroxylase 2 knock-in (Tph2KI) mice. Tph2KI mice exhibit reduced levels of both 5-HT (a) and 5-HIAA (b) in the amygdala compared to WT controls. n = 8 males per group. * p < 0.0005 by t test.

Discussion

Genetic factors, such as mutations in 5-HT-system genes, have been hypothesized to contribute to mental illness by increasing sensitivity to stress. Indeed, 5-HT transporter function has been shown to regulate stress responses in humans (Caspi et al., 2003), rats (Schipper et al., 2011) and mice (Ansorge et al., 2004; Carola et al., 2008). Our present results provide evidence for the importance of 5-HT in regulating responses to early life stress and suggest that the combination of 5-HT deficiency and stress may contribute to impaired behavioural inhibition. These results may have important implications for our understanding of the biological basis for impulsivity-related disorders, such as aggression, drug abuse and suicidal behaviour, the last of which represents a significant and poorly understood problem that has been linked to early life stress (Fergusson et al., 2000), 5-HT deficiency (Shaw et al., 1967) and impulsive-aggressive behaviour (Brent et al., 1994).

Although MS is a relatively mild stress paradigm that has been shown to result in fairly minor anxiety-like behavioural alterations in several mouse lines (Millstein and Holmes, 2007; Parfitt et al., 2007; Savignac et al., 2011), it was sufficient to exert significant effects on multiple behavioural paradigms in our study. It will be interesting to determine whether more severe early life stressors would potentiate the magnitude of the effects observed here or would lead to qualitative differences in stress responses. In addition, future research will be required to investigate the combined effects of 5-HT deficiency and early life stress on other behaviours, including operant impulsivity tasks, aggression and responses to drugs of abuse.

Baseline genotype differences in behavioural assays have the potential to confound interpretations of differential stress responsiveness. For example, given the fact that control Tph2KI mice did not display precautionary grid touches prior to stepping down in the passive avoidance task, the lack of a further impulsivity-promoting effect of MS can be convincingly argued to result from a floor effect rather than from resilience to stress. However, inferring that a floor or ceiling effect is the cause of non-responsiveness in any particular paradigm should be done judiciously as some baseline differences may be further exacerbated by stress.

Given that behavioural disinhibition is considered to reflect a form of impulsivity-like behaviour in rodents (Brunner and Hen, 1997) and most murine indices of anxiety-like behaviour are based on measures of behavioural inhibition (i.e. freezing or avoidance behaviour), increased impulsivity-like behaviour could potentially mask anxiogenic effects. As a result, interpretation of results can be controversial. For example, in examining Tph2 knock-out (Tph2KO) mice, two different groups have reported that Tph2KO animals exhibit a similar phenotype (reduced feeding latency) in the closely related NSF and novelty-induced hypophagia tests. However, one group interpreted this as increased impulsivity-like behaviour, with no effects on anxiety (Angoa-Perez et al., 2012), whereas the other reported this phenotype as decreased anxiety-like behaviour (Mosienko et al., 2012). Similarly, our results in the NSF could be interpreted to indicate that MS is anxiolytic in animals with 5-HT deficiency. However, because of the emergence of behavioural disinhibition in other tests, such as the cliff-avoidance test, which cannot be easily ascribed to reduced anxiety-like behaviour, we favour the interpretation that impaired behavioural inhibition, rather than anxiolysis occurs in Tph2KI mice after MS.

Reports of anxiety-like behaviour using other genetic models of 5-HT deficiency have also been characterized by substantial variability (for review, see Fernandez and Gaspar, 2011). Although the reasons for these differences are obscure, it appears that a genetic × environmental factors interaction may play a crucial role. Indeed, it is possible that differences in the extent of 5-HT deficiency, genetic backgrounds and/or housing conditions could at least partially explain the disparate effects of 5-HT deficiency on anxiety-like behaviour. In addition, it seems likely that some of this controversy results from differences in data interpretation.

Currently, there is considerable controversy regarding the importance of adult hippocampal neurogenesis in the regulation of anxiety-related behaviours (for review, see Eisch and Petrik, 2012). The NSF is one of the primary rodent behavioural paradigms that has been reported to require hippocampal neurogenesis, with increased neurogenesis having been reported to play a role in the reduction in feeding latency induced by antidepressants (Santarelli et al., 2003). Interestingly, Tph2KI mice subjected to MS exhibited reduced feeding latency in this test in the absence of increased levels of hippocampal neurogenesis. However, it is likely the behavioural disinhibition observed in Tph2KI mice after MS in the NSF is mechanistically distinct from the anxiolytic effect of fluoxetine in the same paradigm, and thus these results are not necessarily contradictory. Given the strong link between 5-HT elevation and enhanced neurogenesis (Malberg et al., 2000), it could be considered surprising that low levels of 5-HT did not impair proliferation in the Hip. Regardless, numerous previous studies have reported that inhibiting neurogenesis does not lead to anxiety-like behaviour (Santarelli et al., 2003; Kitamura et al., 2009), although other studies have suggested that hippocampal neurogenesis does play a role in anxiety-like behaviour (Revest et al., 2009; Sah et al., 2012). Nonetheless, given the lack of significant interactions between the effects of 5-HT deficiency and MS on hippocampal cell proliferation and the number of DCX+ immature neurons, it appears unlikely that altered neurogenic responses to MS play a role in the behavioural consequences of 5-HT deficiency following MS.

Adult hippocampal neurogenesis has been shown to play an important role in fear conditioning (Denny et al., 2011), a finding that is consistent with our results demonstrating that MS induces reductions in hippocampal cell proliferation and freezing during fear conditioning. There have been multiple reports of MS leading to decreased hippocampal proliferation in rodents (Mirescu et al., 2004; Kikusui et al., 2009; Lajud et al., 2012), but these effects are also controversial. Indeed, one study reported no effect of MS on hippocampal proliferation but suggested that MS reduced dendritic arborization and neuronal survival (Leslie et al., 2011). In addition, at least one report (using a different early life stress paradigm) has reported increased neurogenesis in response to early life stress (Hays et al., 2012). Our results support the notion that MS exerts a slight negative influence on adult hippocampal proliferation, but we did not find any evidence of altered numbers of immature neurons following MS. Thus, while it is possible that subtle changes in neurogenesis may occur in response to MS in WT and Tph2KI mice, it appears unlikely that differential neurogenic responses to MS underlie the divergent behavioural phenotypes induced by MS in WT and Tph2KI animals.

Similar to our results, previous research has indicated that MS potentiates corticosterone responses to acute stress (Murgatroyd et al., 2009). However, our data, which reveal no significant genotype differences in corticosterone levels under any conditions, do not support a key role for differential corticosterone responses in mediating the differential anxiety- or impulsivity-like responses to MS in WT and Tph2KI mice. Rather, we hypothesize that the stress axis is similarly activated in WT and Tph2KI mice, but that differential responses to stress result from higher neural centres within discrete brain regions, such as the Amyg, which may mediate the altered behavioural responses observed in 5-HT-deficient animals. Indeed, we have previously shown that Tph2KI mice exhibit cortical–amygdalar circuit dysfunction (Dzirasa et al., 2013), a finding that is supported by the aberrant signalling within the Amyg of 5-HT-deficient animals observed here. Lesion studies have implicated the Amyg, together with other regions, such as the NAc, in impulsive-like behaviour (Cardinal et al., 2001; Winstanley et al., 2004). Our data indicate that Tph2KI mice exhibit reductions in 5-HT levels in the Amyg and that MS differentially affects NR1 expression and GSK3β signalling in the Amyg depending on the levels of brain 5-HT. GSK3β has previously been implicated in the regulation of anxiety-like phenotypes (Beaulieu et al., 2008; Mines et al., 2010) and ketamine, which is an NMDA receptor antagonist that has recently demonstrated promise in the treatment of depression (Zarate et al., 2006; Duman and Aghajanian, 2012) and anxiety (Irwin and Iglewicz, 2010). Our results further highlight the potential importance of these signalling pathways in mediating aberrant behaviour and support the notion that targeting these pathways could lead to therapeutic benefits.

Overall, our results indicate that 5-HT deficiency can significantly influence responses to MS, perhaps through modulating signal transduction in the Amyg. These results suggest that 5-HT deficiency may interact with early life stress to determine vulnerability to psychiatric disorders, especially those related to impulsivity. Future research will be required to determine the functional significance of the signalling abnormalities observed in 5-HT-deficient animals after early life stress and to identify novel pharmacologic approaches that reverse the effects of MS.

Acknowledgements

This work was supported in part by grants from the National Institutes of Health (MH79201 and MH60451) to M.G.C. Support from the Lennon Family Foundation to M.G.C. for the initial part of this work is also greatly appreciated. B.D.S. was the recipient of a Minority Supplement award from the National Institutes of Health (MH79201-03S1) and is currently the recipient of an NRSA postdoctoral fellowship (F32- MH093092). W.B.S. was the recipient of an NRSA postdoctoral fellowship (F32-MH083404). A.K. was supported by the Howard Hughes Precollege Program in the Biological Sciences. J.P.R.J. is the grateful recipient of an individual grant from The Lundbeck Foundation of Denmark. We thank Wendy Roberts and Christian Elms for husbandry and care of the mice and Christopher Means and Theodore Rhodes in the Mouse Behavioral and Neuroendocrine Analysis Core Facility for assistance with behavioural testing.

Statement of Interest

M.G.C. has received compensation from Lundbeck as a member of their Psychopharmacology Advisory Board and is a consultant for Omeros Corp. R.M.R. has received compensation from Cleversys Inc., the manufacturer of the behavioural recognition software programs used to analyse behavioural data presented in the manuscript.

Supplementary material

Supplementary material accompanies this paper on the Journal's website.

Supplementary information supplied by authors.

Supplementary information supplied by authors.

Supplementary information supplied by authors.