Serotonin transporter expression is predicted by early life stress and is associated with disinhibited behavior in infant rhesus macaques

Authors


E. L. Kinnally, PhD, Columbia University and New York State Psychiatric Institute, Department of Molecular Imaging and Neuropathology, 1051 Riverside Drive, Room 2917, New York, NY 10032, USA. E-mail: Ek2500@columbia.edu

Abstract

Serotonin transporter (5-HTT) expression patterns may contribute to the risk for adverse psychological outcomes following early life stress. The present study investigated whether two types of early life stress, maternal and social aggression, and a serotonin transporter gene promoter polymorphism (rh5-HTTLPR) predicted lower post-stressor peripheral blood mononuclear cell (PBMC) 5-HTT expression in infant rhesus macaques. We further probed the relationships among these factors and infant behavioral disinhibition within a stressful situation. Fifty-three infants residing with mothers in large, complex social groups were observed over the first 12 postnatal weeks, during which time the rate of aggression received by the infant from their mothers and social group members was recorded. At 90–120 days of age, infants underwent a 25-h maternal separation/biobehavioral assessment, which included standardized behavioral assessments and blood sampling. Infants' rh5-HTTLPR genotypes were determined, and infant 5-HTT expression was quantified from PBMCs collected 8 h after separation. Receipt of aggression from the mother, but not from social group members, was associated with lower post-stressor 5-HTT expression. Lower post-stressor 5-HTT expression, but not receipt of aggression, was associated with disinhibited behavior during assessment. Rh5-HTTLPR genotype was unrelated to any measure. We conclude that 5-HTT regulation is linked with specific, presumably stressful early experiences in infant rhesus macaques. Further, 5-HTT expression predicted behavioral disinhibition, presumably via parallel processes that operate in the brain.

Early life stress in the form of maltreatment or neglect often leads to disadvantageous outcomes, such as psychiatric illness or antisocial behavior (Dodge et al. 1990; Nemeroff 2004), but we have only recently begun to identify the intervening neural and molecular processes (Meaney 2001). The monoamine serotonin (5-HT) system has been strongly implicated in the effects of early life stress on behavior. In particular, the serotonin transporter (5-HTT) regulates reuptake of the 5-HT from the synaptic cleft following release, and therefore plays a key role in 5-HT neurotransmission (Fuller et al. 1991). 5-HTT expression may mediate some of the effects of early adversity, as it is impaired following deprivation or abuse in multiple species (Ichise et al. 2006; Jahng et al. 2007; Lee et al. 2007; Miller et al. 2009) and is lower in depressed (Parsey et al. 2006) and impulsive (Heinz et al. 1998) individuals. Although the influence of early life stress on neurobehavioral outcomes may also depend on structural variation in the 5-HTT gene (rh5-HTTLPR in rhesus macaques: Bennett et al. 2002; 5-HTTLPR in humans: Caspi et al. 2003; Kaufman et al. 2004), little evidence has supported the notion that 5-HTT expression itself is influenced by such gene × environment interactions (Ichise et al. 2006; Kinnally et al. 2008; Miller et al. 2009). The associations of 5-HTT expression with rh5-HTTLPR genotype in the context of species-typical early experiences, however, have not been characterized.

5-HTT availability at a critical developmental period affects emotional behavior throughout the lifespan in mice (Ansorge et al. 2004), indicating that 5-HTT impairment may be linked with neurobehavioral reorganization early in life. Elucidating the regulatory factors of 5-HTT expression and associated behavior early in development is therefore particularly important. Neural or peripheral 5-HTT measures taken at a single point in time have been repeatedly associated with disinhibition/impulsivity in juveniles and adults (gambling, Marazziti et al. 2008; suicide, Roy 1999; impulsivity inventory scores, Coccaro et al. 1996; alcohol sensitivity and aggression, Heinz et al. 1998), but the association between 5-HTT expression and behavior in infancy is unknown. To investigate the role of 5-HTT in the translation between early life stress and impulsive or disinhibited behavior, we measured post-stressor peripheral blood mononuclear cell (PBMC) 5-HTT in vivo in infant rhesus macaques. Peripheral blood mononuclear cells constitutively express 5-HTT and are easily extracted from blood samples (Kinnally et al. 2008). Although it likely does not affect the central nervous system directly, PBMC 5-HTT has previously been linked with neurobehavioral traits (Lima & Urbina 2002) and is dysregulated during stress in maternally deprived infant rhesus macaques (Kinnally et al. 2008). It is presumed that these associations arise from parallel 5-HTT expression between the brain and the periphery (Cupello et al. 2009; Uebelhack et al. 2006). We hypothesized that naturally occurring early life stress (aggression received from the mother and from social group members) was associated with lower post-stressor peripheral 5-HTT expression, consistent with our previous findings in nursery-reared infants. We further predicted that both adversity and 5-HTT expression were linked with behavioral disinhibition in 3–4-month-old rhesus macaques. Finally, we investigated the moderating role of rh5-HTTLPR genotype on 5-HTT expression and behavioral disinhibition in the context of early life stress.

Materials and methods

Experimental subjects

Subjects were 53 (15 males and 38 females) infant rhesus macaques (Macaca mulatta). Infants were raised with their mothers in one of five half-acre outdoor enclosures at the California National Primate Research Center (CNPRC). Each field cage contained a large social group (40–140 members) comprising at least six distinct matrilines with extended kin networks and animals of all age/sex classes. Mean pairwise relatedness between subjects was 3%. Mothers ranged from 3.5 to 17 years of age with a mean of 6.5 years of age.

Procedures

Mother–infant and social behavior observations

Infants were observed in their social groups three times weekly (5 min per observation) between 0700 and 1300 h, during postnatal weeks 1–12. Mother–infant and infant social interactions were coded using a transactional coding system (Lyons et al. 1992), which describes the overall theme of an interaction from the perspectives of the initiator and the recipient. A transaction was defined as a change from one state of association, which had lasted 10 seconds or more, to a new state that is maintained for at least 10 seconds. An aggressive theme was defined as include threatening, biting, chasing, scratching, flattening (pressing into the ground), dragging, pushing away or grabbing infants. Rate of transactions in which the mother initiated aggression or responded to an infant's overture with aggression were calculated. Inter-rater reliability for transactional coding was 85% or better.

Biobehavioral assessment

As part of a large-scale biobehavioral assessment project at the CNPRC, infants (90–120 days of age) were separated from mothers in the field enclosure and transported to an unfamiliar testing suite, and were housed for the next 25 h in individual holding cages (0.81 × 0.61 × 0.66 m) in a temperature-controlled room under a 12:12-h light/dark cycle. Infants experienced a variety of standardized procedures over the 25-h period designed to assess behavioral and physiological reactivity (Capitanio et al. 2006; Golub et al. 2009). Standardized procedures were designed to ensure that each subject had experiences comparable to all other subjects who underwent assessment. A detailed description of these behavioral tests can be found in a previously published report (Capitanio et al. 2006). Briefly, on day 1, animals undergo holding cage observations (0915 h), blood sampling (1100 h), preferential looking test (1130 h), video playback of social conspecifics (1230 h), human intruder observations (1400 h), second blood sampling and dexamethasone injection (1600 h) and second novel object exposure (1630 h). During testing, subjects were allowed access to water and food ad libitum.

Holding cage observation. Five-minute focal observations were conducted on each subject twice during the 25-h separation, and a variety of activity states, emotional behaviors and self-directed behaviors were recorded (see ethogram, Table 1), as described previously (Capitanio et al. 2006). Data collected at 0915 h was included. Novel object interaction. For the duration of day 1 of biobehavioral assessment, a small (5 × 2 × 10 cm) white cylinder was placed in the infant's cage. Inside this object was a recording device (Actiwatch, Philips Respionics, Andover, MA, USA) which is activated whenever a force is exerted on the object. One hour before lights went out, at 1615 h, a new novel object with the same Actiwatch recorder was placed in the holding cage. Data from the recorders were parsed into the number of 15-second intervals for each 5-min period during which any force was exerted on the objects. Force exerted on the object after 1615 h was considered in this analysis. Novel object data was not available for three subjects.

Table 1.  Home cage ethogram
LocomoteDirected movement from one location to another
CrouchVentral surface close to floor; head at or below the level of the shoulders
HangHolding onto ceiling or front mesh; all four limbs off of floor
ActiveWhole body movement; step, jump
ScratchCommon usage
Coo vocalizationMedium-pitched, moderately intense, clear call
Bark vocalizationGruff, abrupt, low-pitched vocalization
LipsmackRapid lip movement usually with pursed lips, accompanied by a smacking sound
ThreatScored with at least two or more of the following: open mouth stare, head bob, ear flaps, bark vocalizations
Fear grimaceExaggerated grin with teeth showing
Tooth grindLoud gnashing of teeth Environmental explore Discrete manipulation by hand or mouth with the physical environment or objects in the cage
EatCommon usage
DrinkCommon usage

Blood sampling and PBMC extraction

Procedures for blood sampling and sample preparation were conducted as described previously (Capitanio et al. 2005; Kinnally et al. 2008). Blood was sampled via femoral venipuncture four times over a 25-h period and each sample was decanted into ethylenediaminetetraacetic acid (EDTA)-treated collection vials. The first sample was collected at 1100 h (AM sample), approximately 2 h following social separation/relocation. The second blood sample was collected approximately 5 h after the first sample at 1600 h (PM sample). For the present study, only PM samples were available for 5-HTT expression analysis.

Whole blood samples were centrifuged for 10 min at 1000 g at 4°C. Plasma was removed and decanted into 1.5-ml Sarstedt (VWR, South Plainfield, NJ, USA) tubes for storage at −80°C. Peripheral blood mononuclear cells were isolated from the remaining sample within 1 h of sampling. White blood cells were aliquoted to RPMI media (Invitrogen, Inc., Carlsbad, CA, USA) supplemented with 10% fetal bovine serum and applied to lymphocyte separation media (MP Biomedicals, Solon, OH, USA). Samples were centrifuged at 800 g at 23°C for 30 min. The purified phase was washed three times with media, centrifuged, and then resuspended in Trizol RNA stabilizing reagent (Invitrogen, Inc.). Samples were stored for no longer than 1 year at −80°C.

Rh5-HTTLPR genotyping

Genotyping was conducted as previously described (Kinnally et al. 2008). Long (l) and short (s) alleles have been described in rhesus macaques (Lesch et al. 1997), and genotype frequencies among our subjects were as follows: l/l (n = 39), l/s (n = 12) and s/s (n = 2). Genotype frequency did not deviate from Hardy–Weinberg equilibrium.

5-HTT expression analysis

5-HTT expression analysis was conducted as described previously (Kinnally et al. 2008). Total RNA was isolated from purified PBMCs stabilized with Trizol reagent (Invitrogen, Inc.). Lysed cells were subjected to phenol extraction and washed with ethanol. Ribonucleic acid was quantified using a spectrophotomer reading at 260/280 nm. One microgram of RNA was then treated with DNAse (Ambion, Inc., Austin, TX, USA) and incubated at 37°C for 60 min. Samples were then subjected to reverse transcriptase polymerase chain reaction (PCR) to synthesize complementary DNA (cDNA). Complementary DNA preparation entailed extension with random hexamers (GE-Amersham Biosciences, Piscataway, NJ, USA), and reverse transcriptase using MMLV-RT (Invitrogen, Inc.). Complementary DNA samples were stored at–20°C for no more than 6 months. Real-time PCR was conducted using an ABI PRISM 7700 Sequence Detection System. A human Taqman quantitative gene expression assay (Applied Biosystems, Inc., Foster City, CA, USA) targeting a region of the 5-HTT gene that we determined to be 100% homologous with rhesus macaques using the published sequence (GenBank accession number AF285761) was used for quantitative PCR. B-actin was selected as an endogenous control, as it was determined to amplify at comparable efficiency to 5-HTT (the slope of the log input amount vs. ΔCt < 0.1). In addition, we confirmed that B-actin did not respond to the stressor or to dexamethasone treatment (all P > 0.05). B-actin probe (Applied Biosystems, Inc.)/primer (Integrated DNA Technologies, Inc., Coralville, IA, USA) sequences are as follows: probe, 5′-ACC ACC ACG GCC GAG CGG-3′; forward primer, 5′-TGA GCG CGG CTA CAG CTT-3′; reverse primer, 5′-CCT TAA TGT CAC ACA CGA TT-3′. A 83.5 ng cDNA was applied to the primer/probe cocktail and Universal Master Mix (Applied Biosystems, Inc.) and run in duplicate. All samples were amplified as following: 2 min 50°C; 10 min denaturation at 95°C, 40 cycles each: 15 seconds at 95°C, 1 min at 60°C. Human RNA (Invitrogen, Inc.) was included on each plate as a control to establish interplate variability. Interplate assay coefficients of variance (calculated as the standard deviation of a human control sample/mean of the human control samples) were less than 3%. 5-HTT values were calculated using the inline image method (5-HTT Ct–B-actin Ct).

Statistics

Maternal and social aggression

The rate of aggressive transactions initiated by the mother toward her infant and the rate of transactions in which the mother responded aggressively to infant-initiated transactions were calculated. Rate data were not normally distributed after several transformation attempts, a common outcome when a large proportion of subjects in an observed population are never observed to display a behavior (approximately 50% in this case). Presence or absence of maternal aggression was therefore coded as a dichotomous variable and used for the final regression models. Analyses that incorporated the rate of maternal aggression as a continuous variable did not differ substantially in outcome from the results presented. The rate of social transactions in which the infant received aggression from other animals (i.e. an adult, juvenile or peer) was also calculated and determined to be normally distributed.

Biobehavioral assessment

Factor analyses, to identify the underlying factor structure in the holding cage assessment behavior, were conducted using M Plus version 5 statistical software, as described elsewhere (Golub et al. 2008). The activity factor included the following behaviors: proportion of time spent locomoting, proportion of time in the hang position (which loaded negatively), rate of environment exploration and whether the animal displayed eating, drinking or crouching. The emotionality factor comprised rates of cooing and barking, and dichotomous codes for scratch, threat and lipsmack. Novel object interaction was calculated as the proportion of twelve 5-min periods during which infants exerted force on the novel object.

Data analysis

Multiple backward regression was used to determine predictors of infant 5-HTT expression and behavior. This backward elimination statistical technique was employed to determine which predictors contributed unique and significant variance to the models and to remove those that did not. The first model included 5-HTT expression as the dependent variable and infant rh5-HTTLPR genotype, presence or absence of maternal aggression, rate of social aggression, the rh5-HTTLPR× social aggression interaction and the rh5-HTTLPR× maternal aggression interaction terms as potential predictors. The second set of models tested the association of these predictors, as well as 5-HTT expression following stress, with three behavioral dimensions (activity, emotionality and rate of novel object interaction) as outcome measures. All statistical analyses were conducted using spss 16.0.

Results

Predictors of 5-HTT expression

The predictive model for 5-HTT expression was significant (F1,52 = 5.391, P = 0.024; adjusted R2 = 0.078). Infant 5-HTT expression was significantly lower in infants that experienced maternal aggression (t = −2.28, df = 52, P = 0.024; Fig. 1), but was not predicted by rate of social aggression directed toward the infant (t = −0.362,df = 52, P = 0.719) or by infant rh5-HTTLPR genotype (t = 0.122,df = 52, P = 0.904). Neither the rh5-HTTLPR× maternal aggression (t = 0.317, df = 52, P = 0.753) nor the rh5-HTTLPR× social aggression (t = −0.477, df = 52, P = 0.636) interaction significantly predicted infant 5-HTT expression.

Figure 1.

Peripheral blood mononuclear cell post-stressor 5-HTT in infants that received aggression or no aggression from mothers in postnatal weeks 1–12.Rh5-HTTLPR genotypes are designated (long homozygotes = l/l; long/short heterozygotes = l/s and short homozygotes = s/s ).

Predictors of infant behavior

The predictive model for infant rate of novel object interaction was significant (F1,49 = 5.302, P = 0.026; adjusted R2 = 0.081). Infant 5-HTT expression was lower in infants that interacted more frequently with a novel object (t = −2.202,df = 49, P = 0.026; see Fig. 2 for plot). Maternal nor social aggression (t = 0.185, df = 49, P = 0.854; t = 0.995,df = 49, P = 0.431), nor infant rh5-HTTLPR genotype (t = 0.468, df = 49, P = 0.642) predicted novel object interaction. Neither the rh5-HTTLPR× maternal aggression (t = 1.038,df = 49, P = 0.305) nor the rh5-HTTLPR× social aggression (t = 1.356,df = 49, P = 0.182) interaction significantly predicted infant novel object interaction.

Figure 2.

Peripheral blood mononuclear cell post-stressor 5-HTT is negatively associated with rate of novel object interaction.Rh5-HTTLPR genotypes are designated (long homozygotes = l/l; long/short heterozygotes = l/s and short homozygotes = s/s ). Linear regression fit lines are illustrated.

The predictive model for infant activity factor scores was significant (F2,52 = 3.207, P = 0.049; adjusted R2 = 0.078). Infant 5-HTT expression was lower in infants that exhibited higher activity scores (t = −2.46, df = 52, P = 0.049; see Fig. 3 for plot). Maternal nor social aggression (t = 0.813, df = 52, P = 0.431; t = 0.820, df = 52, P = 0.416), nor infant rh5-HTTLPR genotype (t = 1.277, df = 52, P = 0.221) predicted infant activity scores. The rh5-HTTLPR× maternal aggression contributed to the significance of the final model, but was not a significant predictor (t = −1.820, df = 52, P = 0.075). The rh5-HTTLPR× social aggression (t = −0.323, df = 52, P = 0.748) interaction did not significantly predict infant activity scores. No predictors of emotional reactivity were identified.

Figure 3.

Peripheral blood mononuclear cell post-stressor 5-HTT is negatively associated with activity factor scores.Rh5-HTTLPR genotypes are designated (long homozygotes = l/l; long/short heterozygotes = l/s and short homozygotes = s/s). Linear regression fit lines are illustrated.

Discussion

Previous studies have suggested that 5-HTT expression plays a role in the translation between early life stress and impulsive behavior. We have extended these findings to show that these links are observable at an early stage in development in rhesus macaques. Using PBMC 5-HTT expression as a model, we confirmed our hypothesis that, within months of birth, infants that experienced aggression from their mothers exhibited lower post-stressor 5-HTT. We also showed that these 5-HTT impairments were associated with behavioral disinhibition (higher rates of novel object interaction and behavioral activity) during a stressful maternal/social separation and relocation at this early stage in development. Neither rh5-HTTLPR genotype nor social aggression, nor genotype × adversity interactions predicted 5-HTT expression or behavior. Although we do not yet know whether post-stressor PBMC 5-HTT mRNA is proportional to either brain 5-HTT mRNA or protein expression, our data are consistent with studies that have associated complex psychological and behavioral processes with peripheral 5-HTT markers (Coccaro et al. 1996; Heinz et al. 1998; Marazziti et al. 2008; Roy 1999). Either peripheral 5-HTT expression following stress is correlated with a distinct neural process that is sensitive to early experience and linked with these behaviors or else it is possible that peripheral 5-HTT-behavior associations arise from correlated 5-HTT expression between the brain and the periphery (platelets; Cupello et al. 2009; Uebelhack et al. 2006). Future studies must confirm that maternal aggression results in impaired 5-HTT expression in neurons as well as PBMCs in rhesus macaques, however. Nonetheless, the present study supports the hypothesis that the associations among early life stress, 5-HTT expression and impulsivity that have been observed in adults (Heinz et al. 1998; Miller et al. 2009) may be rooted in early developmental processes.

Our findings are consistent with a body of literature that suggests that stressful early experiences result in impaired 5-HTT expression (Ichise et al. 2006; Jahng et al. 2007; Kinnally et al. 2008; Lee et al. 2007; Miller et al. 2009). The association of 5-HTT with adversity was specific to the experience of maternal aggression, however, and unrelated to the rate at which an infant experienced social aggression. One might expect that if the reduction in post-stressor 5-HTT was related to the psychosocial stress of aggression experienced during development, that social and maternal aggressiveness would have similar effects. Indeed, rates of aggression from the mother and from social group members occurred at roughly equivalent rates: both averaged about 0.10 incidents per 5-min observation across subjects. We considered the possibility that our experience measures were confounded in their association with infant post-stressor 5-HTT: e.g. mothers may have shifted their behavior according to the infant's social experiences, or infants that received more maternal aggression may have been more likely to incur social aggression. However, observed maternal and social aggression rates were uncorrelated within infants (data not shown).

The specific relationship of maternal aggressiveness to infant post-stressor 5-HTT therefore remains to be explained. Maternal aggressiveness negatively predicted 5-HTT expression even though the average rate of aggression directed toward infants varied greatly: aggression rates directed toward our subjects ranged from low to high (0.07–0.71 incidents per observation period). It is possible that infants were sensitive to incremental variation in maternal behavior due to the singular importance of the mother–infant relationship (Ainsworth 1979; Harlow & Zimmerman 1959; Mason & Mendoza 1998). Because of the developmental importance of this relationship, variation in the quality of mother–infant interactions results in pronounced variation in infant development (Fairbanks & McGuire 1988; Maestripieri et al. 1997; Rosenblum & Paully 1984; Stevenson-Hinde & Simpson 1986). In contrast, it is possible that our maternal aggression measure reflects a more general abusive maternal style. Abusiveness is considered to be a maladaptive maternal behavior pattern and has been defined in macaques as consistent patterns of ‘neglect’ (rejection, refusing infant- initiated suckling), as well as high rates of physical aggression (Maestripieri 1997; McCormack et al. 2006). Indeed, in parallel with our findings, the presumed stress of abuse has been linked with alterations in the 5-HT system (Maestripieri et al. 2006; Miller et al. 2009). It is unlikely that the aggressive mothers in the present study were all abusive; however, we observed approximately 50% of our females to engage in aggression toward infants at some point during the early postpartum period, in contrast to 5–10% incidence of abusiveness in rhesus macaques observed by Maestripieri et al. (2006). We conclude that even relatively small, naturally occurring differences in the experience of maternal aggressiveness are associated with different 5-HTT expression patterns in infants. This finding provides an example of the subtle interplay that occurs between the genome and the environment early in development.

Early adversity impacts neurobehavioral outcomes differently in individuals with structural differences in the regulatory region of the 5-HTT gene (Caspi et al. 2003; Kaufman et al. 2004; although see Risch et al. 2009). We did not find that associations among maternal or social aggressiveness and infant 5-HTT expression or behavior were related to infant rh5-HTTLPR genotype. This is consistent with our previous work [Kinnally et al. 2008; Kinnally et al. (in press)], but inconsistent with other work in rhesus macaques (Barr et al. 2004; Bennett et al. 2002; Champoux et al. 2002; McCormack et al. 2009; Spinelli et al. 2007). In particular, one group has recently shown that infant rhesus macaques that are abused early in life and also possess the low expressing, ‘short’rh5-HTTLPR allele may be at greater risk for physiological and behavioral reactivity to stress (McCormack et al. 2009). Differences between this study and our data may result from differences in maternal behavior measures. As suggested earlier, by the definition of abuse employed by McCormack et al. (2009), it is likely that only a small subset, if any, of our animals would be classified as ‘abused’. Thus, our categorization of early adversity may not have been severe enough to detect underlying gene × environment interactions influencing behavior. Our sample size and/or the influence of other functional variation in the 5-HTT regulatory region (Hranilovic et al. 2004; Hu et al. 2005; Vallender et al. 2008) may also partially explain the inconsistencies between our data and other published studies.

The mechanism(s) of translation of experience to 5-HTT expression patterns remain to be elucidated. It is possible that early stress is associated with reduced expression of transcription factors that bind the 5-HTT regulatory region, which may result in lower post-stressor 5-HTT expression. Our previous work did not provide unequivocal support for the notion that glucocorticoids mediate the effects of nursery rearing on 5-HTT expression, however (Kinnally et al. 2008; although see Glatz et al. 2003). It is also possible that epigenetic mechanisms mediate the effects of experience on 5-HTT expression. Methylation of cytosine–phosphate–guanosine (CpG) islands upstream of the glucocorticoid receptor gene and concurrent histone modifications occurs as a function of experience, with widespread consequences for hypothalamic-pituitary-adrenal axis and behavioral responses to stress (Szyf et al. 2008; Weaver et al. 2004). Similar processes may mediate the effects of early experience on 5-HTT expression described here.

Perhaps most notably, we have shown that post-stressor 5-HTT was associated with infant behavior during a standardized behavioral assessment, independent of maternal, social and genetic factors. Infants that exhibited lower post-stressor 5-HTT were more active and interacted more frequently with a novel object during assessment, a pattern indicative of behavioral disinhibition. Human and non-human primate studies have suggested that locomotion, environmental exploration and engagement of novelty are indicative of impulsivity in situations where effortful inhibition may be desirable (Caspi & Silva 1995; Olson et al. 2002). Further, the association may be relatively specific, as emotional reactivity factor scores were not associated with 5-HTT expression. An important limitation to this conclusion is that our data are correlational, and we are therefore unable to determine the directions of causality among experience, biological and behavioral measures. As our 5-HTT measures were collected after some behavioral observations were made, it is possible that infant activity levels during separation led to individual differences in 5-HTT expression. We theorize, however, because post-stressor 5-HTT was associated with disinhibited behavior both prior to (holding cage activity) and following (novel object interaction) sample collection, that post-stressor PBMC 5-HTT may be a biological trait influenced by early experience, and is associated with behavioral inhibition because it correlates with neural 5-HTT function. This interpretation is consistent with experimental data that has shown that early life stress is associated with limited 5-HTT expression (Ichise et al. 2006; Jahng et al. 2007; Kinnally et al. 2008; Lee et al. 2007; Miller et al. 2009) and that impaired 5-HTT expression is associated with behavioral disinhibition, even when the two measures are obtained at different points in time (Coccaro et al. 1996; Heinz et al. 1998; Marazziti et al. 2008; Roy 1999). We extend these findings to suggest that the dynamics of the peripheral 5-HTT system in response to stress is also associated with behavioral disinhibition, presumably via parallel processes operating in the brain.

Finally, the consequences of early developmental variation in 5-HTT function and behavioral disinhibition in non-human primates are not known. The association between behavioral trait measures collected early in life and later social adaptation and personality in adult non-human primates is only beginning to be understood (Fairbanks et al. 2004; Suomi 2006; Sullivan & Capitanio 2008; Weinstein & Capitanio 2008; Hinde & Capitanio 2009). In humans, behavioral disinhibition measured early in life is a known predictor of impulsivity and aggression in adolescence and adulthood (Caspi & Silva 1995; Olson et al. 2002). If this is also true for non-human primates, our data may support the notion that 5-HTT plays a role in the ‘cycle of violence’ phenomenon observed in humans and rhesus macaques (Caspi et al. 2002; Dodge et al. 1990; Maestripieri et al. 2006), as well as intergenerational transmission of normative psychosocial outcomes.

Acknowledgments

We gratefully acknowledge the assistance of Genesio Karere, Laura Del Rosso, Laura Agostonelli, Linda Fritts, Christine Brennan, Susie Kang and Brett Farnham. This work was funded by NIH RR017584 (L.A.L.), NIH RR019970 (J.P.C.), RR00169 (J.P.C.), National Science Foundation Predoctoral Fellowship (to E.L.K.) and Harry Frank Guggenheim Fellowship (to E.L.K.).

Financial disclosures: The authors report no financial conflicts of interest with the publication of this work.

Ancillary