Early life stress (ELS) increases the risk of the development of depression and anxiety disorders in adulthood (Heim and Nemeroff, 2001). Childhood trauma and neglect exerts a profound influence on behavior and physiology, putting victims at risk for depression, anxiety disorders and substance abuse (de Wilde et al., 1993; Sedlak and Broadhurst, 1996; Agid et al., 2000; Dube et al., 2003). Indeed, ELS may not only increase the risk for developing these disorders later in life, but may also precipitate the onset of these illnesses, increase the likelihood of comorbidity, and impair efficacy of treatment (Friedman et al., 2002; Nemeroff et al., 2003; Gladstone et al., 2004). Adverse early experiences may therefore confer susceptibility to psychiatric disorders in adulthood, in all likelihood, due to the detrimental effects of stressful events on neurodevelopment (Teicher et al., 2003; Watts-English et al., 2006). As the serotonergic system has been identified as crucially important to both the pathophysiology and the treatment of mood disorders, alterations in serotonin neurotransmission following ELS may be one key component in determining the contribution of ELS to depression and anxiety (Owens and Nemeroff, 1994). Although altered serotonin neurotransmission has been reported in subjects with ELS (Rosenblum et al., 1994; Kinnally et al., 2011), the consequences of ELS on serotonin neurotransmission are not well understood and require elucidation. Moreover, large scale studies indicate that less than 30% of patients with major depressive disorder (MDD) remit to the selective serotonin reuptake inhibitor (SSRI), citalopram (Trivedi et al., 2006) prompting renewed scrutiny of the serotonin system. For instance, ELS predicts poor response to treatment with the antidepressant, nefazadone (Nemeroff et al., 2003), suggesting modifications of the serotonin system that negate antidepressant efficacy.

In order to examine the pathophysiological effects of early life adversity, our laboratory has studied the variable foraging demand (VFD) model of ELS in bonnet macaques. In the VFD model, unpredictability of maternal food access compromises maternal care of the infant, producing an experimental homolog of human anxiety and mood disorders in the offspring (Rosenblum and Paully, 1984; Andrews and Rosenblum, 1988, 1991; Rosenblum et al., 1994). Although past research suggests that mood disorders may result from a serotonin deficit in stress-related brain regions, studies of CSF serotonin metabolite concentrations in patients with mood disorders have produced inconsistent results (Shaw et al., 1967; Bourne et al., 1968; Pare et al., 1969; Lloyd et al., 1974). Additional attempts to confirm reduced levels of the serotonin metabolite, 5-hydroxyindoleacetic acid (5-HIAA) in depressed subjects have been inconclusive (Bowers et al., 1969; Papeschi and McClure, 1971; Mendels and Frazer, 1974; Nordström and Åsberg, 1992; Mann et al., 1996a). Researchers have speculated that the inconsistency has been due to CSF 5-HIAA failing to detect localized serotonin deficits in brain sub-regions involved in mood disorders (Mendels and Frazer, 1974). Suicide victims have, however, consistently demonstrated reduced lumbar CSF 5-HIAA concentrations (Mann et al., 1996b). However, regional divergence in regional differences in serotonin and 5-HIAA in suicide victims (Bach et al., 2014) is discussed further below.

In two earlier studies, our laboratory found elevated concentrations of cisternal CSF 5-HIAA in VFD-reared vs. control subjects reared without maternal uncertainty (Coplan et al., 1998; Mathew et al., 2002). The finding of relatively higher concentrations of cisternal CSF 5-HIAA in VFD does not necessarily contradict the notion of a regional serotonin deficit in forebrain regions in mood disorders. We posit that the use of cisternal CSF taps preferentially accesses diffusion of 5-HIAA from the nearby midbrain serotonergic dorsal raphe (DR) region, thus potentially yielding a more accurate picture of localized monoamine activity than that provided by lumbar puncture. The latter conceivably samples an aggregate of serotonin metabolite levels from throughout the CNS, with putative mixing of high midbrain and low forebrain 5-HIAA, thereby obscuring regional differences (Bach et al., 2014). By contrast, a cisternal CSF 5-HIAA deficit is consistently observed in maternally-deprived peer-reared vs. maternally-reared rhesus monkeys (Higley et al., 1991), although it is important to note that the low CSF 5-HIAA associated with peer rearing is linked not only with affective disturbance but also with impulsivity and aggression (Higley and Linnoila, 1997). Using positron emission tomography, peer-reared subjects showed reduced serotonin transporter protein binding in both the dorsal raphe and hippocampus (Ichise et al., 2006), demonstrating that reduced CSF 5-HIAA is associated with deficits of serotonin in regions salient to the current study. Thus, maternal deprivation appears to ostensibly impact the serotonin system differently when compared to maternal unpredictability.

We hypothesized that the high cisternal CSF 5-HIAA found in VFD subjects therefore reflects a specific form of dysfunctional serotonin neurotransmission post-ELS. For instance, suicides are reported to have simultaneously hyperserotonergic midbrain raphe and hyposerotonergic prefrontal cortex (PFC) (Bach et al., 2014). The authors postulated stress-induced increase in tryptophan hydroxylase as the cause of increased brainstem 5-HIAA. PET studies by the same group suggest a greater number of autoreceptors in MDD as a cause of less serotonin neuronal firing and serotonin release (Parsey et al., 2006) but the hypothesis does not incorporate the observation of elevated midbrain 5-HIAA.

During infant neurodevelopment, adverse experiences cause persistent activation of stress-mediated circuits, sensitizing them to exhibit exaggerated responses to future stress (Heim and Nemeroff, 2001; Artigas et al., 2006). Previously identified glutamatergic DR afferents (Valentino et al., 2003) which arise in layer V of the medial prefrontal cortex, a stress-sensitive region (Celada et al., 2001), may induce heightened release of extracellular serotonin in the midbrain dorsal raphe region, activating a negative feedback mechanism mediated by 5-HT_1A autoreceptors. An alternate hypothesis invokes a developmental increase following ELS in midbrain tryptophan hydroxylase and 5-HIAA (Wilson et al., 2012). An overall reduction in serotonin neurotransmission to limbic structures such as the hippocampus, and other neurotrophic targets such as related white matter tracts, may ensue, independent of hypothesis (Celada et al., 2001; Valentino et al., 2003). We therefore surmise that high cisternal CSF 5-HIAA concentrations may be directly linked to an attenuation of serotonin outflow (Blier et al., 1998). Subsequently, this mechanism may represent an important facet of the increased vulnerability to stress and mood disorders following ELS (Heim and Nemeroff, 2001; Coplan et al., 2011).

In the current study we seek to replicate our finding of elevated CSF 5-HIAA in a third cohort of VFD animals. Furthermore, we sought to confirm our hypothesis that high cisternal CSF 5-HIAA is associated with a putatively reduced serotonin transmission from the DRN by first correlating the former with hippocampal volume. An inferred attenuation in serotonin output to the hippocampus would have a directly negative effect on serotonin intracellular messaging systems and neurotrophic activity. Diminished numbers of neurotrophic factors specifically affects neuronal growth and survival, leading to reduction in hippocampal volume (Duman et al., 1997, 2000). Dopamine and norepinephrine metabolites were also examined [homovanillic acid (HVA) and 3-methoxy-4-hydroxyphenylglycol respectively]. We have previously demonstrated elevated CSF HVA in VFD-reared subjects vs. non-VFD controls (Coplan et al., 1998), implicating dopamine as another monoamine system altered by early life stress. Although we failed to demonstrate rearing differences for CSF MHPG, we had previously reported altered sensitivity to the alpha-2 receptor antagonist, yohimbine, in VFD- vs. non-VFD reared subjects (Rosenblum et al., 1994). We wished to investigate whether there were alterations in the relationship between CSF MHPG and hippocampal volume following ELS, as reflective, in part, of serotonin-related alterations. We reported that in patients with panic disorder, enhanced serotonin neurotransmission induced by SSRI administration is associated with lowered plasma MHPG concentrations (Coplan et al., 1997). Serotonin attenuates neuronal response to glutamate, a major neurotransmitter utilized by afferents to the noradrenergic locus ceruleus (LC) (Aston-Jones et al., 1991). Reduced outflow of serotonin to target projections such as LC, resulting from elevations of peri-raphe serotonin, would favor CSF MHPG positively tracking CSF 5-HIAA. Thus, the relationship between CSF MHPG and hippocampal volume would be expected to mirror the relationship between CSF 5-HIAA and hippocampal volume.

Although less documentation exists, compromised serotonin neurotransmission may well influence the neurotrophic status of other brain regions in addition to the hippocampus. Alterations in white matter (WM) integrity as measured by FA have been observed throughout the brain across a range of psychiatric disorders (Daniels et al., 2013). Based on the notion that serotonin-dependent neurotrophic growth factors affect WM integrity measures (Martinowich and Lu, 2007), our own studies previously found FA reductions following ELS in the anterior limb of the internal capsule (ALIC), which lies in key circuitry connecting the temporal and frontal regions of the brain (Cannistraro et al., 2007; Zou et al., 2008). Although the uncinate fasciculus has featured prominently in studies of childhood maltreatment (Huang et al., 2012), and the fornix is closely related to the hippocampus (Kubicki et al., 2005), the ALIC shows decreased FA in neurodevelopmental disorders such as schizophrenia (Kubicki et al., 2005) and schizotypal personality disorders (Hazlett et al., 2012). Decreased FA of the ALIC has also been identified as a plausible link between type II diabetes and major depressive disorder (Zhang et al., 2013). In the present study, we examine the relationship between cisternal CSF 5-HIAA concentrations and FA measures derived from the ALIC. We hypothesized that high cisternal CSF 5-HIAA, specifically in VFD-reared subjects, would also be inversely associated with WM integrity in the ALIC, providing convergent evidence for serotonin-related neurotrophic compromise following ELS.

The primary hypothesis, therefore, was to detect elevated CSF 5-HIAA in VFD vs. non-VFD. The second hypothesis was to demonstrate an inverse correlation between CSF 5-HIAA concentrations and hippocampal volume in VFD, indirectly invoking a putative deficit in hippocampal serotonin neurotransmission. The third hypothesis was to demonstrate an inverse relationship between CSF 5-HIAA and FA of the ALIC in VFD, providing convergent evidence of neurotrophic compromise in serotonin target areas.

The current study replicates our finding of elevated cisternal CSF 5-HIAA in two previous cohorts of macaque subjects reared under the VFD model of ELS (Mendels and Frazer, 1974; Coplan et al., 1998). The data indicate that maternal uncertainty produced by variable foraging conditions may render offspring vulnerable to developing abnormally elevated levels of CSF 5-HIAA (Mendels and Frazer, 1974; Coplan et al., 1998). Effects are independent of age, weight and sex, which is consistent with our previous studies (Coplan et al., 1996; Mathew et al., 2002) but in contrast to rhesus studies, where age, pubertal status and sex have been demonstrated to influence CSF 5-HIAA concentrations (Higley et al., 1991). In this VFD model of ELS, we find that putatively high peri-raphe concentrations of serotonin are correlated with markers of reduced neurotrophic activity in structures affected by serotonin pathways, specifically the right hippocampus, suggesting that high cisternal CSF 5-HIAA is associated with attenuated neurotrophic effects.

To date, lumbar CSF 5-HIAA measures to characterize the status of serotonin neurotransmission in mood disorders, with the exception of suicidal patients, have reported inconsistent findings (Bowers et al., 1969; Papeschi and McClure, 1971; Mendels and Frazer, 1974). These inconsistencies may stem, in part, from sampling of CSF via lumbar puncture. Lumbar CSF may reflect an aggregate of spinal, midbrain, frontal, and limbic projection-site 5-HIAA concentrations, therefore failing to identify monoamine metabolite measures in specific localized brain regions. The concentration gradient of monoamine metabolites in lumbar CSF means levels vary with the amount of fluid taken (Johansson and Roos, 1975; Sjöström et al., 1975), suggesting differences in regional concentrations. In macaques, cisternal CSF is usually drawn from the cerebello-pontine angle via an atlanto-occipital route reasonably adjacent to the midbrain, which may preferentially reflect diffusion from peri-raphe extracellular fluid. Regional differences, however, in CSF 5-HIAA concentrations have not, to our knowledge, been systematically reported, and CSF 5-HIAA may represent a homogenous pooled resource coming from multiple areas of the brain. For instance, Ichise et al. (2006) reported a strong correlation between serotonin transporter binding and cisternal CSF 5-HIAA concentrations in a number of areas of the brain, but interestingly did not find a correlation between transporter binding in the raphe and CSF 5-HIAA concentrations.

Our model, however, is in accordance with a recent pilot study of suicides that found high levels of 5-HIAA in the midbrain (and was also associated with reduced 5-HIAA in projection tissues such as the PFC) (Bach et al., 2014). In past studies, we have demonstrated that VFD subjects exhibit elevated CSF corticotropin-releasing factor (CRF) measures and these CSF CRF concentration elevations correlated directly with CSF 5-HIAA concentrations only in VFD-reared subjects (Coplan et al., 1996). These heightened levels of CRF reveal a hyperactive stress-response system in the VFD brain (Holmes et al., 2003a), a scenario found to persist long term in rodents subjected to early life stress and human patients with a perception of early adversity (Liu and Weiss, 2002; Holmes et al., 2003a; Carpenter et al., 2004; Huot et al., 2004; Ladd et al., 2004; Brunson et al., 2005). The current findings are intriguing given growing evidence of a close inter-relationship between CRF and serotonin in mediating stress responses (Roche et al., 2003). Recent data show that both stress and CRF release attenuate serotonin neurotransmission in specific forebrain structures, suggesting that stress may activate CRF-mediated inhibitory afferents to the DRN, decreasing serotonin release to the forebrain (Roche et al., 2003). In addition, excitatory glutamate-mediated mPFC afferents stimulate serotonin release in the DRN in response to stress, in turn activating a negative feedback mechanism via inhibitory somatodendritic 5-HT_1A autoreceptors located on serotonergic neurons in the DRN, attenuating serotonin neurotransmission (Blier et al., 1998; Celada et al., 2001; Valentino et al., 2003). Adult knockout mice lacking the gene for the serotonin transporter protein (Brunson et al., 2005) exhibit increased concentrations of peri-raphe serotonin compared to wild-type mice, providing a useful model for how the serotonin system functions under conditions of high dorsal raphe serotonin. In these subjects, the dorsal raphe serotonergic neurons are in a perpetual state of hyperpolarization (Lira et al., 2003), with diminished synaptic serotonin at terminal projection sites, such as the hippocampus. Behaviorally, both depression-like and anxious behaviors were observed in these serotonin transporter knockout mice (Lira et al., 2003; Holmes et al., 2003b). Taken together, the data bolster the plausibility that elevated cisternal CSF 5-HIAA may in fact be indicative of a serotonin deficit downstream of the raphe nucleus, but the study was not designed to document this putative serotonin deficit. Moreover, we have hypothesized elsewhere that elevations of midbrain serotonin would counteract the efficacy of SSRI's, as appreciable 5-HT_1A autoreceptor downmodulation, essential for SSRI efficacy (Blier et al., 1987), would no longer be feasible (Coplan et al., 2014).

In the current study, we found that relatively high VFD cisternal 5-HIAA measures correlated with reduced right hippocampal volume, a potential index of neurotrophic activity. Hippocampal volume is closely reliant on the presence of neurotrophic factors such as brain-derived growth factor (BDNF) to maintain neuronal growth and survival (Duman et al., 2000). The production of BDNF and other similar factors is instigated by serotonin-initiated cellular messenger signal proteins, implying that reduced serotonin neurotransmission at the hippocampus may conceivably cause reduced volume (Duman et al., 2000; Daniels et al., 2013). Other signaling pathways may also increase BDNF such as peripheral growth factors that are increased by physical exercise (Cotman et al., 2007). BDNF, nevertheless, has an important synergistic relationship with the serotonin system, regulating serotonergic neuronal survival while serotonin-dependent second messengers modulate BDNF production (Martinowich and Lu, 2007). The hippocampus is particularly vulnerable to changes in BDNF level; missense polymorphisms in the BDNF gene are directly correlated with reduced hippocampal volume (Szeszko et al., 2005).

That reduction of FA measures of white matter integrity in the ALIC inversely associated with elevated cisternal CSF 5-HIAA provide convergent validation of neurotrophic compromise secondary to a putative serotonin deficit state in downstream structures of the serotonergic pathway. Parental verbal abuse is linked to a decrease in FA in a variety of white matter tracts (Choi et al., 2009), indicating there is a link between ELS and WM integrity. A number of studies illustrate the influence of serotonin on WM integrity. Variations in WM integrity appear to be regulated by serotonin transporter genotype in rhesus monkeys (Howell et al., 2014), while patient response to SSRIs is predicted by FA measures (DeLorenzo et al., 2013). Intracellularly, the serotonin-dependent neurotrophin, BDNF, enhances WM integrity by increasing concentrations of myelin basic protein and facilitating oligodendrocyte proliferation (McTigue et al., 2001; Wong et al., 2013). The current study reveals that relatively elevated cisternal CSF 5-HIAA is associated with reduced FA in the ALIC of VFD-reared subjects. We acknowledge that other areas of white matter in addition to the anterior limb of the internal capsule such as the fornix, which is intimately related to the hippocampus (Kuroki et al., 2006), should have been considered for study. We contend that the choice of the ALIC is justified, not only because reduced FA in the ALIC has been found in a range of psychiatric disorders (Kubicki et al., 2005; Zhang et al., 2013) but that our contention is that the high cisternal CSF 5-HIAA compromises serotonin-related function across a range of target projection site, and that the ALIC provides a useful WM proxy of neurotrophic compromise. Further supporting our results is the finding of an inverse relationship between the dopamine and norepinephrine metabolites—HVA and MHPG, with hippocampal volume in VFD subjects. These associations are statistically absent in non-VFD subjects. CSF HVA and 5-HIAA levels consistently show a close correlation (Papeschi and McClure, 1971). With respect to MHPG, we have previously reported that in patients with panic disorder, enhanced serotonin neurotransmission by SSRI's is associated with lowered plasma MHPG concentrations (Coplan et al., 1997). In the noradrenergic locus ceruleus (LC), serotonin attenuates neuronal response to glutamate, a major neurotransmitter utilized by afferents to the LC (Aston-Jones et al., 1991). In the VFD midbrain, elevated cisternal CSF 5-HIAA, acting via inhibitory 5-HT_1A autoreceptors, may conceivably attenuate serotonin outflow, thereby reducing serotonin-mediated inhibition of LC neuronal firing (Aston-Jones et al., 1991). In this putative serotonin deficit state, paradoxically high CSF 5-HIAA would be positively associated with CSF MHPG. Thus, inverse associations between cisternal CSF HVA and MHPG concentrations and right hippocampal volume serve as proxy indicators of the relation between cisternal CSF 5-HIAA and hippocampal volume, specifically in subjects reared under conditions of ELS.

Although our findings seemingly conflict with the CSF 5-HIAA deficit consistently observed in maternally-separated peer-reared vs. maternally-reared rhesus monkeys (Higley et al., 1991), it is important to note that the low cisternal CSF 5-HIAA associated with peer rearing is linked not only to affective distress (Clarke and Snipes, 1998) but also impulsivity and aggression (Higley et al., 1991; Higley and Linnoila, 1997), whereas VFD subjects exhibit consistent disturbances in affective domains without impulsivity and aggression (Rosenblum and Paully, 1984; Higley et al., 1991; Rosenblum et al., 2001). Thus, the two forms of ELS—maternal deprivation vs. maternal uncertainty—result in divergent CSF 5-HIAA findings and, despite overlap in the affective domain, divergence in impulsive and aggressive behavior.

Moreover, in vervet monkeys, reduced cisternal CSF 5HIAA was also found to be inversely associated with social impulsivity (Fairbanks et al., 2001). By contrast, subordinate female rhesus, a chronic form of unpredictable stress, is associated with high CSF 5-HIAA (Asher et al., 2013). The data imply that specific forms of ELS may induce specific variations in serotonin neurotransmission. A range of distinct neurobiological sequelae in response to various patterns of ELS have therefore been reported (Mathew et al., 2001).

Limitations of the present study include the restriction of the imaging data to a subset of the males in the sample although the CSF 5-HIAA elevations following ELS were not sex-specific. The inverse association between CSF 5-HIAA, MHPG, and HVA in VFD-reared subjects and hippocampal volumes was restricted to the right side. It is interesting to note that we have previously reported in this same cohort reduced left hippocampal volume in VFD-reared vs. non-VFD-reared controls whereas differences for right hippocampal volume were not significant (Jackowski et al., 2011). It is conceivable that the left hippocampal volume has been reduced by ELS to an extent that a “floor” effect is in place. Thus, left hippocampus is not sufficiently intact to be affected by altered ranges of serotonin neurotransmission. A major limitation of the paper is that a serotonin deficit state in target projection areas can only be inferred and was not directly documented. Future post-mortem studies on VFD vs. control subjects may provide a methodology whereby a serotonin-deficit state in VFD hippocampus can be definitively ascertained. Nevertheless, precedent for high peri-raphe serotonin and/or its primary metabolite, 5-HIAA, and low PFC serotonin/5-HIAA is present in suicides (Bach et al., 2014). Moreover, a preclinical precedent for low hippocampal serotonin in the face of high peri-raphe serotonin is observed in serotonin transporter knockout mice (Lira et al., 2003). Moreover, the acute effects of serotonin transporter blockade by SSRIs is to increase peri-raphe serotonin, reduce serotonin neuronal firing and reduce distal synaptic release, negating immediate antidepressant effects of the SSRIs (Haddjeri et al., 1998). Despite cogent precedents of high peri-raphe serotonin leading to reduced serotonin outflow, direct documentation of a hippocampal serotonin deficit state in VFD subjects with high cisternal CSF 5-HIAA is required in order to draw a definitive conclusion. Although we have posited that cisternal CSF is a preferable site to document alterations of serotonin neurotransmission, going forward, direct quantification on post-mortem samples of non-human primates would be required to provide definitive validation of the view that ELS effects an increase in midbrain serotonin, thereby inducing a serotonin deficit state in terminal fields. For instance, Knott et al. (1989), reported a positive relationship between ventricular 5-HIAA and cortical 5-HIAA in post-mortem brain of patients with Alzheimer's disease, emphasizing the need for regional quantification. The use of the serotonin metabolite, 5-HIAA, requires validation as a marker of 5-HT. Increased turnover, which may reduce serotonin and increase 5-HIAA, should be ruled out by direct measurement of serotonin. A further limitation of the study is the age difference between the VFD-reared vs. normally-reared controls. Although this difference is only in the order of six months, one primate year is approximately 3–4 human years. This would suggest that the 6 month difference between VFD and controls of 6 months could be an 18 months to two year difference in humans. However, despite this limitation, using age as a covariate did not affect the significance of the results. Furthermore, age was not related to CSF 5-HIAA, either when considering male and females combined, males alone or females alone. In addition, the subjects in the current study were at an age where they were peripubertal (Kaufman and Rosenblum, 1966). Puberty has been noted to affect CSF 5-HIAA in rhesus (Higley et al., 1991). Lack of measurement of gonadal hormones prevents covariation for the confounding effects of pubertal status. Furthermore, since females enter puberty before males, the fact that the VFD group were 6 months older than the non-VFD could further compound the confound introduced by gonadal hormones. Using weight as a proxy for pubertal growth, we noted that there were no weight effects on CSF 5-HIAA, either when considering male and females combined, males alone or females alone. Moreover, there were no sex effects for weight and there was a trend for a rearing group × sex effect for body weight. Post-hoc t-tests revealed that VFD females were not heavier than non-VFD females but because this comparison was a near trend, weight was used as a covariate when assessing rearing group differences for CSF 5-HIAA. Elevations in CSF 5-HIAA concentrations in VFD compared to non-VFD subjects remained significant. Finally, the inverse relationship between CSF 5-HIAA and right hippocampal volume is correlational but not causal. Other factors, such as HPA axis alterations, may represent an intermediary step between alterations in serotonin function and hippocampal volume. Moreover, without causal evidence, the directionality of the hypothesis stemming from altered serotonin function to adversely affect hippocampal remains an assumption. Evidence for the hypothesis, to the extent supported by convergent processes, is provided by the inverse relationship between CSF 5-HIAA concentrations and white matter FA. Studies directed at confirming the causality of CSF 5-HIAA elevations in the pathophysiology of hippocampal volume reductions are warranted.