Experience of stress in childhood negatively correlates with plasma oxytocin concentration in adult men

Abstract

Early life experience is known to affect responses to stress in adulthood. Adverse experience in childhood and/or adolescence sensitises to life events that precipitate depression in later life. Published evidence suggests a relationship between depression and oxytocin (OT), but the extent to which early life experience influences OT disposition in adulthood deserves further exploration. This study hypothesised that early life stress (ELS) has a long-term negative effect on OT system activity. The study was performed on 90 male volunteers (18–56 years; mean ± standard deviation = 27.7 ± 7.09 years). Several questionnaires were used to assess: health, early life stressful experiences in childhood (ELS-C, up to 12 years) and early life stressful adolescence (13–18 years), recent stressful life events, depressive symptoms, state–trait anxiety and social desirability. Plasma OT concentration was estimated by means of a competitive enzyme immunoassay. Lower OT concentrations were significantly associated with higher levels of ELS-C (p < 0.01), and with depressive symptoms and trait anxiety (both p < 0.05). The interaction between ELS-C and trait anxiety was significant (p < 0.05), indicating that the link between ELS-C and plasma OT concentration is moderated by trait anxiety. These results contribute to the evidence that early life adverse experience is negatively associated with OT system activity in adulthood, and offer further insight into mediator and moderator effects on this link.

Keywords::

• Anxiety
• coping
• depression
• HPA axis
• oxytocin
• stress

Introduction

There is consensus that early life adverse experience, such as infant/child abuse or parental neglect, represents a risk factor in the aetiology of mood disorders including depression (McEwen 2000; Heim and Nemeroff 2001; Shea et al. 2004). It has been suggested that childhood abuse or neglect sensitises to stress in later life (Harkness et al. 2006) in agreement with the stress sensitisation hypothesis (Post 1992). Consequently, childhood maltreatment has been suggested to be a vulnerability factor that should be assessed in diagnosis and treatment of depression (Harkness et al. 2006). A recent study of the long-term effects of sexual abuse in childhood and adolescence indicates that such early life experience, particularly when unresolved, increases perception of acute psychosocial stress (Pierrehumbert et al. 2009). This is linked with the flattening of cortisol responses in adult women, thus distorting the normal psychobiological reactivity to stress (Pierrehumbert et al. 2009). Early life stress (ELS) seems to trigger a cascade of neurobiological effects that result in brain remodelling, as observed in adult humans and animals. Thus, magnetic resonance imaging studies have demonstrated a decrease in the hippocampal volume in adult women with a history of childhood abuse and a diagnosis of post-traumatic stress disorder (Stein et al. 1997) or major depression (Vythilingam et al. 2002). Preclinical studies on rodent and non-human primate models of ELS show that early life maternal deprivation leads to the development of anxiety and depression-like behavioural phenotypes, with reduced motivation for reward and reduced coping in adulthood (Sánchez et al. 2001; Pryce et al. 2005). These behavioural changes are linked with long-term neurobiological alterations in the limbic system, namely reductions in astroglial density and morphology (Leventopoulos et al. 2007; Opacka-Juffry et al. 2008) and a decrease in 5-HT1A receptor binding in rats (Leventopoulos et al. 2009). Whilst the primary neurobiological mechanism driving these effects remains unknown, the involvement of the hypothalamus–pituitary–adrenal (HPA) axis is plausible (Pryce et al. 2002; Shea et al. 2004). Early experience also affects the oxytocin (OT) and vasopressin systems in animals in a gender-dependent manner (Carter et al. 2009). Such changes can alter HPA-axis reactivity to stress, because OT and vasopressin, the active elements of the hypothalamo-neurohypophysial system (HNS), modulate stress-induced responsiveness of the HPA axis at multiple levels; this includes potentiation of adrenocorticotropic hormone (ACTH) secretion by vasopressin released from axons of hypothalamic parvocellular and magnocellular neurones (Engelmann et al. 2004). Whilst vasopressin stimulates ACTH secretion from the anterior pituitary, OT tends to reduce it (Legros 2001).

OT is well known for its roles in sexual functions, such as ejaculation, and in parturition, milk ejection and maternal behaviour in animals and humans (Gimpl and Fahrenholz 2001). It is a neuropeptide synthesised within the hypothalamic paraventricular and supraoptic nuclei (PVN and SON, respectively) and released into blood in the posterior pituitary (as part of the HNS), providing peripheral (circulating) OT. OT is also present and active in the brain, within the central pathways where OT receptors are expressed (Gimpl and Fahrenholz 2001). The relationship between circulating (peripheral) OT and its brain presence is of relevance to studies where blood OT is analysed, particularly because circulating OT does not effectively cross the blood–brain barrier (1–2% of a dose given peripherally passes to the brain; Uvnäs-Moberg 1998) and OT release into the circulation does not necessarily reflect its release into the brain (intracerebral). There are many lines of evidence that OT released intracerebrally coordinates maternal behaviour, mother–infant bonding and affiliative behaviour, the latter in both female and male animals (Ross and Young 2009). In male rodents, intracerebral OT regulates sexual behaviour (Argiolas and Gessa 1991) and reduces anxiety when administered intracerebroventricularly (Ring et al. 2006) or into the PVN (Blume et al. 2008) or released endogenously in the PVN (Waldherr and Neumann 2007). Anxiolytic effects of intracerebroventricularly administered OT were also observed in female rats (Windle et al. 1997). In humans, centrally administered OT (intranasal application) enhances trust, effects of social support, positive social memories and reduction of aggression and anxiety (Heinrichs et al. 2003; Huber et al. 2005; Kosfeld et al. 2005; Guastella et al. 2008; Heinrichs and Domes 2008).

Many researchers assume dissociation between central and peripheral OT release and action, which is consistent with the view that separate neuronal populations support the posterior pituitary OT system and the brain projections, with magnocellular neurones of PVN and SON sending axons to the pituitary and parvocellular PVN neurones projecting centrally to the brain (Ross and Young 2009). OT can be released from dendrites of magnocellular OT neurones and it has been proposed that OT released somatodendritically within the hypothalamic nuclei can diffuse to distant brain regions that express OT receptors (Ludwig and Leng 2006; Leng and Ludwig 2008). Interestingly, a recent neuroanatomy study on prairie voles and rats (Ross et al. 2009) demonstrates that both central and posterior pituitary (peripheral) OT projections originate from magnocellular neurones, which argues for coupling between intracerebral and posterior pituitary OT release. This is also consistent with the earlier functional observations that stressful conditions of forced swimming trigger an increase in extracellular release of OT in the PVN and SON synchronised with OT release into the circulation in rats (Wotjak et al. 1998).

Consistent with the involvement of the HNS in HPA-axis regulation (Engelmann et al. 2004), stress-related clinical conditions like depression seem to implicate OT. For example, patients with major depression exhibit a negative correlation between peripheral OT level and the scored symptoms of depression and anxiety in women (Scantamburlo et al. 2007). By contrast, Parker et al. (2010) reported increases in plasma OT level and no changes in cortisol in 11 patients (males and females) with major depression.

As there is growing evidence from both experimental and clinical studies that OT is involved in the control by the brain of responses to stress and stress-related conditions in animals and humans (Uvnäs-Moberg 1998; Legros 2001; Engelmann et al. 2004; Heinrichs and Gaab 2007; Scantamburlo et al. 2007; Neumann 2008; Parker et al. 2010), the link between ELS and OT in later life is seen as clinically relevant (Carter et al. 2009). Experimental studies on rodents indicate that early life experience can have long-term effects on the OT and vasopressin systems, through changes in expression of their central receptors (Carter et al. 2009). In primates, pathogenic rearing conditions for Rhesus monkeys result in abnormal affiliative behaviour and stress responses to novel environment, which are associated with significantly reduced cerebrospinal fluid (CSF) concentrations of OT in adult animals, when comparing nursery-reared with mother-reared conditions (Winslow et al. 2003). Notably, in humans, early parental separation associates with attenuated responses (decreases) of salivary cortisol to exogenous OT (nasal spray) as shown in a study on a small sample of young male volunteers (Meinlschmidt and Heim 2007), whilst history of abuse in childhood associates with lower CSF concentrations of OT in women (Heim et al. 2009).

Taking into account the above, we aimed to investigate the link between plasma OT and ELS in a large psychiatrically non-referred male adult sample. Despite the evidence that several types of maltreatment often co-occur (Rosenberg 1987), previous studies have focussed on the experience of abuse (emotional, physical and sexual) and neglect (emotional and physical; Fries et al. 2005; Heim et al. 2009). Therefore, in the present study, we aimed to measure a broader range of ELS, i.e. experience of loss and illnesses in both childhood (up to 12 years) and adolescence (13–18 years). Previous research has shown that plasma OT concentrations can be either negatively or positively correlated with the symptoms of depression and anxiety (Scantamburlo et al. 2007; Parker et al. 2010) and that ELS is positively correlated with such symptoms (Kendall-Tackett et al. 1993). Thus, it is important to explore whether the link between ELS and OT is mediated by the symptoms of depression or trait anxiety. We also aimed to investigate whether social desirability and state anxiety could influence the link between ELS and OT.

We hypothesised that ELS would negatively relate to plasma OT concentrations such that high levels of ELS would associate with low concentrations of OT, and that the effects of ELS on OT concentrations would be stronger for individuals high in depressive symptoms or high in trait anxiety.

Materials and methods

The project was approved by the Roehampton University Ethics Board in March 2008.

Participants

The study was performed on 98 healthy male volunteers who were recruited through advertisements. The participants were fully informed of the purpose of the study, signed an informed consent form and received £20 as an incentive. The following exclusion criteria were applied: regular heavy exercise, medical conditions (e.g. heart disease, diabetes, hypertension), current or previous clinical psychosomatic or psychiatric diseases, allergies, atopic diathesis, rheumatic diseases, HIV, infectious or blood diseases, smoking, alcohol intake of >10 units/day, recreational drugs, medication, poor sleep pattern, body/mass index (BMI) of >32 kg/m² and a tendency to faint. The volunteers were instructed not to eat or drink for at least 1 h before the beginning of the study. They were allowed to relax and answer the questionnaires prior to venepuncture. They had been informed about the taking of blood samples and asked not to volunteer if they had a fear of bleeding.

Seven participants were excluded due to incomplete questionnaire data or due to haemolysis in the plasma samples used for OT analysis. The remaining 90 participants (93% of the initial sample) had a mean age of 27.70 years standard deviation (SD = 7.9; range 18–56), 85% were native English-language speakers and had a mean BMI of 24.4 (SD = 3.06, range 17.4–30.3). In terms of educational attainment, 37.6% of participants were educated to university degree level, 51.1% to college level and 11.1% to secondary school level only; 65.6% of participants described themselves as employed or self-employed, 17.8% were students and 16.7% were unemployed. Age and BMI did not correlate with any other variables. As there were no differences between the group of native English speakers and the remaining participants or between the occupational status or educational attainment and any other variables, these variables were not considered any further.

Measures

Questionnaires

Early life stressful experiences in childhood (ELS-C, up to 12 years) and early life stressful adolescence (ELS-A, 13–18 years) were measured using Early Life Stress Inventory (ELSI, Mohiyeddini and Opacka-Juffry, in preparation; supplementary Table). ELSI is a 60-item (30 identical items for each life stage) self-report inventory which provides retrospective assessment of the frequency of different types of stressful life experiences considered as uncontrollable or fateful (e.g. abuse, neglect or loss of friend or close relative). As it has been known that the accuracy of recall of childhood experience is higher for concrete events rather than the subjective experience of events (Brewin et al. 1993), the items in ELSI are phrased in terms of concrete, objective experiences or events (e.g. loss of sibling) or behaviours (e.g. serious arguments with parents) in order to maximise the accuracy of recall. The ELS-C and ELS-A inventories demonstrated an internal consistency (using the Kuder-Richardson Formula 20 for dichotomous items, see Cortina 1993) of 0.87 and 0.86 when validated. Furthermore, the ELS-C inventory showed a test–retest reliability in a sample of 186 non-clinical adults with values ranging from 0.82 to 0.90 over a period of 5–6 months, indicating resistance to reporting biases caused by transient mood or memory states. ELSI showed a high convergent validity with a clinician-rated interview of childhood and adolescence, and therapists' ratings of critical life events in a sample of 87 clinical patients. In a non-referred sample of 92 students (mean age = 22.6 years, SD = 1.5), the monotrait–heterosource correlations between separately measured self-reported ELS-C and ELS-A and mother-reported ELS-C and ELS-A were r(89) = 0.58 (p < 0.001) and r(89) = 0.69 (p < 0.001), respectively. Overall, the above demonstrates that ELSI is a reliable and valid self-reported measure of critical life events in childhood and adolescence.

Recent stressful life events (RSLEs) within the last 12 months were measured using a social readjustment rating scale (Holmes and Rahe 1967). RSLE is the most frequently used measurement of life stress. It assigns numbers, called life-change units, to 43 critical life events with their life-change value, ranging from 100 (death of spouse) to 11 (minor violations of the law). The life-change values are then summed to yield a total score that indicates how much “stress” the individuals had experienced. Scores < 150 reflect a slight risk of illness, 150–299 reflect a moderate risk and >300 reflect a serious risk.

Depressive symptoms were measured according to the Center for Epidemiological Studies-Depression Scale (CES-D; Radloff 1977). The CES-D is a widely used 20-item self-report for determining depressive symptomatology and is intended for the general population (Radloff 1977). We used a 4-point response format ranging from 0 (rarely or none of the time) to 3 (all of the time). Cronbach's α was 0.90 for this subscale.

State and trait anxiety subscales (each consisting of 20 items) of the Spielberger's State–Trait Anxiety Inventory (Spielberger et al. 1983) were used to measure trait and state anxiety. The trait subscale had a 4-point response format ranging from 1 (almost never) to 4 (almost always). The state subscale had a 4-point response format ranging from 1 (not at all) to 4 (very much so). Cronbach's α was 0.90 for each subscale.

In order to measure one form of response bias or “faking good”, social desirability was measured using the short version (Strahan and Gerbasi 1972) of the Marlowe–Crowne Social Desirability Scale (MCSDS; Crowne and Marlowe 1964). This short version consists of 10 true–false statements in which five items describe socially approved but infrequent behaviours (e.g. “I am always willing to admit it when I make a mistake”), and the other five items refer to socially disapproved but frequent behaviours (e.g. “I like to gossip at times”). The items had a 2-point response format ranging from 0 (false) to 1 (true). A high score on this scale reflects an individual tendency towards faking good. Cronbach's α was 0.62 for the short version of MCSDS.

In order to investigate whether the fear of venepuncture had an impact on plasma OT or the recall of ELS, a short “Fear of Bleeding Scale” (FBS) with three questions (Are you afraid of blood collection procedure? Is blood collection procedure unpleasant for you? Are you worried about blood collection procedure?) was developed. The items had a 5-point response format ranging from 0 (very slightly or not at all) to 4 (extremely). Cronbach's α for FBS was 0.92. However, FBS was not correlated with any other variable and thus was not considered any further.

Oxytocin analysis

Venous blood samples (2 ml) were collected into chilled EDTA (1 mg/ml) tubes between 14:00 and 17:00 h; one sample was collected per subject. They were centrifuged at 4°C and the plasma samples were kept chilled until frozen in two aliquots within 1 h of collection. The plasma samples were stored at − 80°C until analysis within 5 months of collection. The person performing the assay was blind to the subject status of the samples.

OT was analysed in duplicate unextracted plasma samples by enzyme immunoassay (Assay Designs, Ann Arbor, MI, USA) in the clinical research laboratory of Roehampton University. Plasma samples that showed high OT concentrations (33% total) were diluted up to five times and re-assayed using the second of two stored plasma aliquots to avoid thaw–freeze artefacts. Solid-phase extraction of the plasma samples was not carried out in order to avoid potential artefacts caused by varied extraction rates. No inconsistencies, indicative of matrix effects, were observed when assessing the effects of plasma dilutions on OT concentrations. Intra-assay precision was 8.1 and 8.9% for medium and high OT concentrations, their mean values being 91.2 and 309.5 pg/ml, respectively; inter-assay precision was 4.8 and 10.5% for medium and high OT concentrations, respectively. The sensitivity of the enzyme immunoassay was 11.7 pg/ml (range 15.6–1000 pg/ml).

Statistical analyses

All calculations were performed using SPSS Inc. (version 16) software packages (SPSS, Chicago, IL, USA). The obtained data are presented as mean ± SEM. The optimal sample size of n = 95 to detect an effect size of η² = 0.25 (representing a small effect size) with a power 0.80 or greater and α = 0.05 was calculated a priori with the statistical software G-Power (Buchner et al. 1997). In cases of missing data, the results were excluded listwise. Data were tested for normal distribution using the Kolmogorov–Smirnov test before the statistical procedures were applied. Log transformation was applied to correct the positive skew in the OT data. The association between OT values and the psychological data was described by computing Pearson's correlation coefficients (two tailed). The mediation and moderation functions of depressive symptoms or trait anxiety were investigated using multiple regression analysis (see Baron and Kenny 1986; Frazier et al. 2004). Results were considered statistically significant at p ≤ 0.05. Using raw OT concentration values, the means and SDs were calculated for each participant. Data of one participant were excluded, because his duplicate OT values were more than 3 SDs greater than the group mean. Additionally, four participants were excluded due to haemolysis in plasma samples.

Results

Table I displays the descriptive statistics and the associations between the variables.

Table I.  Descriptive data for measures (means, SEs and bivariate correlations).

	1	2	3	4	5	6	7	8
1. Oxytocin (log transformation)	–
2. Early life stress-childhood	− 0.34**
3. Early life stress-adolescence	0.02	0.57**
4. Recent stressful life events	− 0.01	0.25*	0.41**
5. Depressive symptoms	− 0.23*	0.38**	0.42**	0.23*
6. Trait anxiety	− 0.22*	0.20	0.20	0.09	0.37**
7. State anxiety	− 0.09	− 0.01	− 0.06	− 0.08	0.13	0.48**
8. Social desirability	0.07	− 0.05	− 0.20	0.02	− 0.27*	− 0.27*	− 0.20	–
Mean	5.72^†	4.66	3.98	314.32	0.83	2.21	2.12	0.55
SE	0.06	0.50	0.48	28.08	0.07	0.06	0.06	0.02

Note: *p < 0.05; **p < 0.01 (Pearson's correlation coefficients); SE, standard error

^† “oxytocin” expressed in pg/ml.

The mean OT plasma concentrations were 377.6 ± 23.9 pg/ml, and the range was 78.6–1198pg/ml. The most frequent stressful experiences in childhood were (see Supplementary Table) a loss of friend or close relative and a change of school. The most frequent stressful experiences in adolescence were serious arguments with parents and being a victim of crime. The Kolmogorov–Smirnov test revealed that ELS-C and ELS-A were normally distributed (Zs < 1.32, ns). The Kuder-Richardson Formula 20 for dichotomous items indicated an internal consistency of 0.89 and 0.87 for ELS-C and ELS-A, respectively.

Trait anxiety, state anxiety and depressive symptoms were normally distributed. In contrast, the distribution of depressive symptoms was not normal (Z = 1.70, p < 0.01), consistent with a large proportion of participants having lower scores on this scale. However, when considering scores >27 as being high on the scale of depressive symptoms (Boyd et al. 1982; Zich et al. 1990), 17.8% of the present participants fell within the range of depressive symptoms despite the fact that those men reported no clinical diagnosis at any point in their lives.

Exploratory correlational analyses (Table I) showed that lower plasma concentrations of OT were significantly associated with higher levels of ELS-C, depressive symptoms and trait anxiety (all p < 0.05). Furthermore, a hierarchical regression analysis (with depressive symptoms, trait anxiety and state anxiety entered in Step 1 and ELS-C entered in Step 2) revealed that the link between ELS-C and plasma OT concentration remained significant (β = − 0.29, t(85) = 2.63, ΔR² = 0.07, p < 0.05) and accounted for variance above and beyond depressive symptoms and anxiety. In contrast, the link between plasma OT and trait anxiety, and plasma OT and depressive symptoms vanished after controlling the impact of ELS-C.

ELS-C and ELS-A were significantly correlated (r(90) = 0.57, p < 0.001). However, OT concentrations were not correlated with ELS-A, RSLE, age, social desirability or state anxiety (rs(90) < 0.08, ns). In addition, ELS-C was associated with RSLE and with depressive symptoms (both p < 0.05). ELS-A was associated with depressive symptoms and with RSLE (both p < 0.001). RSLE was positively associated with depressive symptoms (p < 0.05). Depressive symptoms and trait anxiety were positively associated and showed negative correlations with social desirability (both p < 0.01). State anxiety was associated with trait anxiety (p < 0.001).

Exploratory mediator analysis assessed whether the link between ELS-C and OT concentration was mediated by depressive symptoms or trait anxiety. According to Baron and Kenny (1986), four conditions must be met for a variable function to be a mediator variable: (1) the link between the predictor (ELS-C) and the outcome variable (OT concentration) must be significant, (2) the link between the independent variable and the mediator variable (depressive symptoms or trait anxiety) must be significant, (3) the link between the mediator and the outcome variable must be significant and (4) the mediator variable has to reduce (partial mediation) or eliminate (total mediation) the link between the predictor and the outcome variable. Here, Condition 2 was not fulfilled for trait anxiety and hence trait anxiety could not mediate between ELS-C and OT. Although the correlations between ELS-C, OT concentration and depression support Conditions 1, 2 and 3, Condition 4 was investigated by means of multiple regression analysis, following Baron and Kenny (1986). The present results revealed that ELS-C had a substantial impact on OT concentrations (β = − 0.30, t(88) = 2.74, p < 0.001). In contrast, the impact of depressive symptoms predicting OT concentration was not significant (β = − 0.12, t(88) = 1.11, ns). Hence, the depressive symptoms did not constitute a link between ELS-C and OT in the present sample.

Exploratory moderator analysis assessed whether depressive symptoms or trait anxiety alters the direction or strength of the relation between ELS-C and OT. For each of the moderator variables (depressive symptoms and trait anxiety), hierarchical multiple regressions were computed. In each set of hierarchical regressions, the ELS-C and the moderator variable were entered first in the equation, followed by the interaction term (Table II). Although there were no signs of multi-collinearity (the variance inflation factor (VIF) values were below 2.5 and tolerance statistics were all well above 0.2), all variables were standardised in order to equate different metrics used by measuring variables under investigation (Dunlap and Kemery 1987). The interaction terms were created as product terms (ELS-C × trait anxiety and ELS-C × depressive symptoms).

Table II.  Testing moderator effects of depressive symptoms and trait anxiety.

Step and variables	B	SE B	95% CI	β	R^2	Adjust R^2
Depressive symptoms as moderator
Step 1
ELS-C	− 0.17	0.07	− 0.31, − 0.04	− 0.28*
Depressive symptoms	− 0.08	0.07	− 0.21, 0.05	− 0.13	0.12**	0.12**
Step 2
ELS-C × depressive symptoms	0.09	0.06	− 0.04, 0.22	0.15	0.02
Trait anxiety as moderator
Step 1
ELS-C	− 0.18	0.06	− 0.31, − 0.06	− 0.30**
Trait anxiety	− 0.10	0.06	− 0.23, 0.02	− 0.17	0.14**	0.16**
Step 2
ELS-C × trait anxiety	0.12	0.06	0.00, 0.24	0.21*	0.04*

Note. *p < 0.05; **p < 0.01 (multiple regression analysis). CI, confidence interval and ELS-C, early life stress-childhood.

The results revealed a significant interaction effect (Figure 1) between OT concentrations and ELS-C × trait anxiety that indicates that trait anxiety does function as a moderator of the relationship between ELS-C and OT concentrations (β = 0.24, t(86) = 2.38, ΔR² = 5%, p < 0.05). The moderation function of trait anxiety remained significant even after including depressive symptoms and BMI in the first equation and controlling its associations with variables under investigation. Accordingly, the impact of ELS-C on OT concentration changed linearly with respect to trait anxiety; the negative effect of ELS-C on OT increased gradually and steadily as the positive effect of low trait anxiety decreased. After controlling for the first-order effects, the interaction between ELS-C and trait anxiety explained the 5.3% of incremental variance in OT concentrations (p < 0.05). Figure 1 summarises the moderator effects for those subjects who scored at the mean ± SD on the ELS-C and trait anxiety.

Figure 1.  Summary of a moderator effect: stressful life events in childhood (ELS-C), trait anxiety and plasma OT. Trait anxiety moderates the link between ELS-C and OT (n = 85; β = 0.24, t(86) = 2.38, p < 0.05, multiple regression analysis). In order to facilitate the interpretation of the graphs, log-transformed OT values are expressed as corresponding to the mean (mean ± SD) of ELS-C frequency and trait-anxiety scores.

Furthermore, exploratory analysis was carried out on the correlation between ELS-C and ELS-A, and an overall score was computed averaging ELS-C and ELS-A in order to cover the time window up to 18 years. This score was not correlated with OT (r(90) = − 0.19, ns). However, it was significantly correlated with symptoms of depression (r(90) = 0.45, p < 0.001) and trait anxiety (r(90) = 0.23, p < 0.05).

Discussion

The present exploratory study investigated a relationship between ELS and plasma OT concentrations in male adults with no psychiatric history (as self-reported). The ELS–Childhood and Adolescence Inventories (ELSI; Mohiyeddini and Opacka-Juffry, in preparation) were used as a retrospective measure of the frequency of different types of stress (validation explained in the Methods section). The most frequent childhood stress experiences were change of school, loss of friend or close relative indicating that it is important to measure a broader range of ELS beyond abuse and neglect in psychiatrically normal adults. The core finding of this study is a negative association between ELS in childhood and peripheral OT levels in adulthood in men. Plasma OT concentrations were significantly associated with higher levels of ELS-C, depressive symptoms and trait anxiety. In addition, moderator analysis demonstrated that trait anxiety moderates the link between ELS-C and OT.

The present OT concentrations with a mean value of 378 pg/ml appear to be high when compared with most of the literature data on human plasma OT (Pierrehumbert et al. 2010, mean values of 7.14 and 20.67 pg/ml in two parallel sets of extracted plasma analysed by means of radioimmunoassay, RIA). However, they are comparable with some others (Zak et al. 2005, mean value of ∼200 pg/ml in their “random condition”, as measured by means of ELISA with no plasma extraction) or even considerably lower (Scantamburlo et al. 2007, mean OT plasma concentrations 3.67 ng/ml, i.e. 3670 pg/ml in patients with major depression, as measured by RIA in unextracted samples). It is possible that unextracted plasma gives higher values in the course of immunoassay, through some unknown matrix interactions, which is a limitation even though the reproducibility of the present method is good. The present procedure is similar to that used by Zak et al. (2005).

There can be many psychobiological implications of the present finding, considering that the OT systems in the brain are involved in the regulation of emotional behaviour (Neumann 2008), development of social interactions (Heinrichs et al. 2003; Guastella et al. 2009), reactivity to stress (Uvnäs-Moberg 1998; Legros 2001) and building and maintaining resilience (Ozbay et al. 2008). However, as discussed above, measurement of plasma OT concentration is not necessarily an indicator of OT release in the brain. There are limited published studies on a link between early life experience and OT activity in humans. Of these, a study by Fries et al. (2005) on children, who were placed immediately after birth in orphanages and suffered early life neglect, concludes that despite a longer period of successful adoption, those children had lower urinary concentrations of OT after interactions with their mothers than their age-matched controls from a typical home environment. However, that work has been criticised for methodological reasons (Anderson 2006) and does not provide psychometric analysis. Only recently Heim et al. (2009) reported lower levels of OT in the CSF of women with a history of abuse in childhood. That important contribution was derived from a small sample of 22 female volunteers. By contrast, a new study by Pierrehumbert et al. (2010) found that childhood abuse was associated with higher plasma OT levels in adult women who were exposed to an acute psychosocial test (Trier Social Stress Test), which can be interpreted as a sign of dysregulation of the normal reactivity to stress. The present results are consistent with the findings regarding CSF OT by Heim et al. (2009), whilst offering data from a larger group of 90 male volunteers.

When comparing research data concerning OT in males and females, one should be aware of gender effects on the peripheral and central OT systems (Gimpl and Fahrenholz 2001) as well as gender differences in responses to stress, with “flight or fight” being proposed as an adaptive male reaction whereas “tend and befriend” as a female response to stress (Taylor et al. 2000). OT responses to stress are also gender dependent as confirmed by experimental studies which demonstrate that male rats exhibit lesser posterior pituitary hormone response to stress than females (Williams et al. 1985). Estrogen increases OT receptor binding and gene expression (Young et al. 1997), whilst early life experience exerts gender-dependent effects on stress reactivity (Wigger and Neumann 1999) and the OT and vasopressin systems in animals (Carter et al. 2009).

A noteworthy element of the present study is the lack of significant association between stressful life events in adolescence and recent stress and OT plasma levels. It is intriguing that the present results indicate different effects of childhood stressful experiences versus adolescent stressful experiences on OT plasma levels in adulthood, taking into account that adolescence is considered to be a period of enhanced neurobehavioural vulnerability due to ongoing brain remodelling (Sisk and Foster 2004). It is possible that sex hormones play a role because of their regulatory effect on the HPA axis as has been documented in animal studies; this is of particular significance in adolescence (McCormick and Mathews 2007). Adolescent males appear to be less sensitive than females to stressors that augment social instability, such as isolation or novelty of environment, as derived from experimental studies (McCormick et al. 2004).

The present study offers further interesting findings. Firstly, the results of partial correlation analysis show that the link between ELS-C and OT is independent of depressive symptoms and trait anxiety although a considerable proportion of participants reported high levels of depressive symptoms. Secondly, the link between ELS-C and OT is not mediated by trait anxiety or depression, indicating that these three factors represent different aspects of the pathophysiological process. Thirdly, the significant interaction between ELS-C and trait anxiety reveals that as ELS-C increases, the positive effect of low trait anxiety on OT levels decreases. Notably, this interaction is independent of individual levels of depressive symptoms or BMI. It shows that simultaneously higher scores of ELS-C and trait anxiety are associated with lower OT concentrations, beyond that accounted for by the main effects of ELS-C and trait anxiety. Although the differences between R² and the adjusted R² are small (which indicates that a good cross-validity of these models can be assumed), it would be crucial to cross-validate this result empirically in future research considering the size of this effect.

Limitations

There are several shortcomings of this cross-sectional exploratory study, which is restricted to males and correlational. Correlation does not allow for deriving causal inferences on a relationship between ELS and adult OT activity. Also, the sample was self-reported as normal, i.e. without psychiatric diagnosis (e.g. structured clinical interview for DSM). We used a self-reported measure of ELS, which like any retrospective self-report may be affected by reappraisal (Fivush 1993) by redefinition of past events (Ross 1989), by impairment of memory (Squire 1989; Bremner 1999) or by reporting biases. In addition, although the present results show that fear of venepuncture had no impact on plasma OT, emotional states or ELS recall, a prior venous catheterisation would have been more appropriate as it would have allowed for collecting multiple blood samples, thus possibly smoothing a pulsatile OT baseline (Cyranowski et al. 2008). Even though every effort was made to keep the times of plasma handling the same and minimal for all the samples, it is recognised that OT stability in plasma is limited. OT was analysed in plasma samples without prior extraction, which could potentially lead to matrix effects in the assay as discussed above. Finally, given the non-experimental design and a relatively small sample size, the power of tests detecting true interaction effects should be considered as low (Aguinis 1995).

Conclusions

The present results derived from a sample of men with no reported psychiatric diagnosis indicate that ELS may affect OT system activity in the long term, whilst the baseline behavioural trait of relevance to mood disorders remains unchanged. It is conceivable that such a neurohormonal shift can facilitate increased sensitivity and/or reduced resistance to repetitive stressors, e.g. psychosocial, which in turn can trigger mood disorder, including first episodes of depression. This is of clinical relevance, and future studies should consider whether patients with recurrent depression have lower plasma concentrations of OT in association with a history of childhood stress experience.

Acknowledgements

This study was supported by School of Human and Life Sciences, Roehampton University, UK (grant number P276). The authors wish to thank Mr Balbir Josen Singh and Mr Donald Fisher for their technical assistance. Preliminary results of this study were presented to the British Neuroscience Association conference (April 2009), Harrogate, UK.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.