|Home | About | Journals | Submit | Contact Us | Français|
We tested the hypothesis that early life stress would persistently compromise neuronal viability of the hippocampus of the grown nonhuman primate. Neuronal viability was assessed through ascertainment of N-acetyl aspartate (NAA) – an amino acid considered reflective of neuronal density/functional integrity – using in vivo proton magnetic resonance spectroscopic imaging (MRSI). The subjects reported herein represent a re-analysis of a sample of nineteen adult male bonnet macaques that had been reared in infancy under induced stress by maternal variable foraging demand (VFD) (N = 10) or control rearing conditions (N = 9). The MRSI spectral readings were recorded using a GE 1.5 Tesla machine under anesthesia. Relative NAA values were derived using NAA as numerator and both choline (Cho) or creatine (Cr) as denominators. Left medial temporal lobe (MTL) NAA/Cho but not NAA/Cr was decreased in VFD subjects versus controls. An MTL NAA/Cho ratio deficit remained significant when controlling for multiple confounding variables. Regression analyses suggested that the NAA/Choline finding was due to independently low left NAA and high left choline. Right MTL showed no rearing effects for NAA, but right NAA was positively related to body mass, irrespective of denominator. The current data indicate that decreased left MTL NAA/Cho may reflect low neuronal viability of the hippocampus following early life stress in VFD-reared versus normally-reared subjects. Given the importance of the hippocampus in stress-mediated toxicity, validation of these data using absolute quantification is suggested and correlative neurohistological studies of hippocampus are warranted.
Study of the neurobiological basis of environment-gene interaction during neurodevelopment is fundamental to our understanding of the origins of psychiatric disorders (Bennett et al., 2002; Carpenter et al., 2004; Uher and McGuffin, 2010). Environmental contributions to psychopathology are perhaps most influential during infancy, when neurodevelopmental vulnerability to stressors is particularly manifest. In this study we examined the neurobiological basis of an important environmental factor – early life stress.
Studies of early-life stress in humans can be confounded by multiple factors. Using animal models enables us to control the characteristics of early-life stress and to ensure a normalized environment thereafter (Rosenblum and Paully, 1984). Moreover, longitudinal evaluation is facilitated by the shorter life span of bonnet macaques as compared to humans. In addition, the genetic similarity between humans and macaques along with a structural and functional parallel in their central nervous systems make bonnet macaques well suited subjects to model human behavior and pathology. In our model of early-life stress, although adequate food is always available, macaque mothers face the uncertainty of food availability through a procedure dubbed “variable foraging demand” for 16 weeks within the first year of infant life (Rosenblum and Paully, 1984).
Variable foraging demand (VFD) rearing is a mother-infant early-life stress paradigm in which bonnet macaques are reared by mothers undergoing an experimentally-induced sustained “perception” of food uncertainty. The unpredictability of foraging conditions prevented the mother from attending to the needs of the infant, thereby, disrupting normative patterns of maternal rearing and infant attachment (Coplan et al., 2005). Subsequently, VFD-offspring even as adults manifest a range of behavioral and neurobiological alterations (including neuropeptide, somatotropic axis, monoaminergic system, metabolic syndrome profile, insulin resistance, HPA axis, and cytokines) (Coplan, 2009). These bio-behavioral abnormalities reflect aspects of compromised affective regulation that have been identified in humans in association with the vulnerability, emergence and persistence of mood and anxiety disorders (Andrews and Rosenblum, 1991; Coplan et al., 1996). Current data provide a rationale for testing the hypothesis that the VFD form of early life stress induces a persistent neurochemical milieu that may lead to long-term impairment of neuronal viability of the hippocampus (Law et al., 2009; Mathew et al., 2001). Magnetic resonance spectroscopy imaging (MRSI) provides regional in vivo assessment of the magnitude of N-acetyl-aspartate (NAA) resonance signal within the primate medial temporal lobe (MTL), a region directly corresponding to hippocampus.
NAA is a widely distributed amino acid believed to be a marker of neuronal density, integrity, and/or function (Bertolino et al., 1997). NAA reductions have been interpreted as reflecting neurotoxicity, neurodegeneration, or impaired mitochondrial energy metabolism (Clark, 1998). Additional functions of NAA have been identified (Barker, 2001), most notably its purported role as an osmolyte or water molecular “pump” (Baslow, 2003). Whereas regional reductions in NAA have been argued to represent an in vivo signal of compromised neuronal integrity or fitness (Barker, 2001; Bertolino et al., 1997; Bertolino et al., 2002), accurate interpretation of MTL NAA ratio reductions in nonhuman primates is complicated by multiple confounding variables, requiring consideration of factors such as neurometabolite denominators, body mass and laterality effects. In assessing NAA using MRSI, choline (Cho) and Creatine (Cr) are used as denominators in NAA ratio measures in an effort to provide purportedly more stable NAA ratio measures (e.g., (Schuff et al., 1997)). Cho increases have been interpreted to reflect increased cell membrane turnover (Duyn et al., 1993) and Cr increases have been associated with increased myelination (Kreis et al., 1993). In addition, Cr is involved in energy-dependent brain function (Wyss and Kaddurah-Daouk, 2000). Cr decreases are presumably associated with increase in metabolic activity (as discussed in Coplan et al., 2006), As one example, Cr has reportedly been reduced in medial temporal lobe of patients with panic disorder (Massana et al., 2002), raising the possibility that Cr may also vary independently in pathological anxiety states. Increases of baseline NAA ratio measures in right dorsolateral prefrontal cortex of human patients with generalized anxiety disorder (Mathew et al., 2004) and Asperger Syndrome (Murphy et al., 2002), provide precedent for relatively high NAA states, perhaps following excess cleavage of N-acetyl aspartatyl-glutamate (NAAG) into glutamate and NAA.
In disorders such as multiple sclerosis, Alzheimer’s disease, and epilepsy, reduced NAA is correlated with the degree of neuronal and/or axonal damage or loss (Barker, 2001; Cheng et al., 2002; Petroff et al., 2002). In contrast, in psychiatric disorders such as schizophrenia and bipolar disorder, in which cell loss appears relatively minor, reductions in prefrontal cortical and hippocampal NAA may, instead, represent a state of neuronal compromise (Arnold and Rioux, 2001). Therefore, in psychiatric disorders the hippocampal neurometabolite abnormalities may be more evident as a marker of hippocampal alterations than morphometric abnormalities (Bremner et al., 1997; Bremner et al., 2000). For example, in one study, post-traumatic stress disorder (PTSD) patients exhibited approximately 30% less hippocampal NAA than controls despite there being only a 5% hippocampal volume difference between the groups ((Schuff et al., 2001); also see (Freeman et al., 1998)).
Other human MRSI studies of psychiatric disorders are generally concordant with the view that an increase in NAA signal may be an in vivo marker of restoration or repair of neuronal viability. Moore et al. (2000) demonstrated that lithium administration increased whole-brain NAA in humans, thereby suggesting an augmentation of neuronal viability. Furthermore, the increased brain NAA correlated with volumetric increases in gray matter regional volume. In a second treatment study, Michael et al. (2003) reported an increase in left periamygdalar NAA in depressed patients who had responded to electroconvulsive therapy.
Conversely, proton MR spectroscopy measures were used in humans to observe long-term post-irradiation metabolic changes in epileptogenic hippocampal tissue (Hajek et al., 2003). Short-term irradiation has been used to suppress neurogenesis while sparing mature neurons (Santarelli et al., 2003). Changes in the MTL in patients with epilepsy, following radiosurgery on the amygdala and hippocampus, were assessed using MRSI for up to 3 years. Follow-up was characterized by a decrease of NAA, Cr, and Cho concentrations ipsilateral to the irradiation and was significantly decreased from ipsilateral control values and concentrations in the contralateral part of the brain. No radiotoxic changes were observed (Hajek et al., 2003). Chronic psychosocial stress in tree-shrews leads to reductions in whole brain voxel NAA as well as dentate gyrus neurogenesis, an effect reversed by the antidepressant, tianeptine (Czeh et al., 2001). Thus, the findings of these two studies suggest that a decrease in NAA levels may under certain circumstances serve as a marker of reduced neurogenesis in the dentate gyrus.
Reflecting the hypothesized homologous relationship of VFD to components of human anxiety and mood disorders, recent studies in our laboratory have found that adult male VFD-reared macaques exhibit decreased NAA/Cr in a voxel combining right and left anterior cingulate cortex (ACC) compared to normally reared controls (i.e., low foraging demand; LFD) (Mathew et al., 2003). The ACC is critically implicated in multiple forebrain functions including emotional regulation and mediating executive function (Allman et al., 2001). In addition, the ACC is a major recipient of hippocampal efferent outputs and comprises one of the major components of the circuit of Papez (Martin et al., 2003). Thus, a discrete period of stressful early rearing resulted in persistent compromise of neurometabolic fitness or integrity (or decreased number of neurons) within the ACC, a neighboring direct projection site of the hippocampus (Martin et al., 2003).
In this study, we extended MRSI examination of ACC NAA concentrations to include separate analyses of left and right hippocampi on the same subjects. In the original Mathew et al. (2003) report using NAA/Cr and NAA/Cho, there were no significant group differences when hippocampus values were merged from right and left side. However, brain laterality data in humans and nonhuman primates link positive emotional states with relative left hemispheric activation, while negative emotional states have been linked with relative right hemispheric activation (Kalin et al., 2000; Miskovic et al., 2009; Schore, 2002). Accordingly, based on the translational clinical data and the persistent affective disturbances seen in VFD-reared subjects (Rosenblum and Paully, 1984; Rosenblum et al., 2001), it would be hypothesized that adverse early rearing might preferentially attenuate left MTL neurometabolic status. In addition, the relationship between NAA and confounding constitutional factors, such as age and body mass, had not been analyzed separately for each hippocampus.
Of note, there were no mean differences between the groups for age or weight, suggesting that paired matching had been effective in controlling for these two independent variables. Of the four designated NAA dependent variables, only left hippocampal NAA/Cho demonstrated significant decreases in VFD subjects (N = 7) versus controls (N = 6), Bonferroni corrected (see Figure 1 for scatterplot; see Table 1 for statistics). In addition, effects remained significant when controlling for body weight and Cho/Cr. No differences were noted in right hippocampal NAA/Cho, or in NAA/Cr values from either side.
Using VFD rearing as an outcome variable and left hippocampal NAA/Cho and Cho/Cr as input variables, bivariate regression analyses indicated that low left NAA/Cho in contrast to high left Cho/Cr significantly correlated with VFD rearing. The two variables combined contributed to almost 50% of the variance of VFD rearing. For right hippocampus, relatively high Cho/Cr exhibited a near significant predictive effect for rearing status but right NAA/Cho was not predictive of rearing status (for statistics see Table 2). The overall variance accounting for VFD rearing by these latter two right hippocampal neurometabolite ratios did not achieve significance. In summary, these analyses suggest a left sided NAA deficit, independent of the high left Cho/Cr VFD rearing differences evident on ANOVA.
In order to further ascertain the relative contributions of individual neurometabolites to left MTL variance, regression analyses were simultaneously performed on two ratio measures where a common denominator was present, presumably accounting for the identical denominator variance independent of numerator used. The next analysis used ratio values for NAA/Cho and Cr/Cho [derived by inverting Cho/Cr (1/neurometabolite value)]. Thus, the relative contribution of NAA versus Cr to VFD outcome was determined while controlling for Cho as a common denominator. Low left MTL NAA/Cho contributed significantly whereas low left MTL Cr/Cho contributed at a trend level to over 50% of VFD rearing variance. In addition, using left MTL NAA/Cr and Cho/Cr in prediction of rearing status, both low left NAA/Cr and high left Cho/Cr were each significant predictors in opposite directions of rearing group differences. The combination of the two independent sources of variance -- low NAA and high choline -- contributed to over 50% of overall rearing variance. Therefore, using a common denominator control approach, evidence was provided that low left MTL NAA was evident irrespective of denominator used.
Evidence for significantly lateralized hippocampal neurometabolic function was notable when the statistical contribution of neurometabolite ratios to predicting body mass variance was examined, an analysis for which the two rearing groups were combined. Prior to combining groups, in non-VFD, body mass correlated significantly with right hippocampal NAA/Cho (r = .91, N =7, p = .005) and NAA/Cr (r = .8; N =7; p <.03). In VFD subjects, body mass showed a trend for correlation with NAA/Cho (r=.62; N=10; p = .056) and NAA/Cr (r = .53; N=10; p = .11). For right hippocampus using subjects from both groups, NAA/Cho associated strongly with body mass whereas right Cho/Cr did not associate with body mass. Neither left hippocampal NAA/Cho nor left Cho/Cr ratios significantly contributed to body mass variance. A high portion of body mass variance could be accounted for by both the right hippocampal NAA/Cho and NAA/Cr ratios. In contrast, for left hippocampus, combining NAA/Cho and NAA/Cr ratios accounted for minimal variance.
Body mass and age were highly co-linear (see Table 3). In contrast to age, body mass was more homogenously distributed across the 19 subjects studied and was therefore used for parametric correlation comparisons. A positive association between body weight and NAA signal was markedly evident in the right MTL region. The association was positive independent of denominator (both NAA/Cr and NAA/Cho associations were significant (see Table 3). In contrast to left hippocampus, rearing group had no discernable effect on either right hippocampal NAA ratio. Left hippocampus differed significantly from right hippocampus in its relationship between NAA to Cho, using Cr as the mutual denominator. In left hippocampus, a strong correlation suggested a close positive relation between neuronal fitness and membrane turnover requirements. In contrast, in right hippocampus, NAA bore no relationship to Cho. Right hippocampal NAA/Cr and NAA/Cho were more strongly correlated with body weight respectively in comparison to correspondent left MTL NAA ratios (Table 3).
The current data indicate that low left MTL NAA/Cho may constitute a persistent spectral marker, potentially associated with low neuronal viability of hippocampus in VFD-reared versus normally-reared subjects. Moreover, regression analyses suggest that relatively low left NAA, low left Cr and high left Cho each contribute independent significant or trend sources of variance for left MTL in the VFD condition. Regression analyses exhibited low left NAA when controlling for high choline, suggesting the absence of rearing group effects for NAA/Creatine may be an artifact of simultaneous reduction of both NAA and Cr neurometabolites in VFD MTL (considering the findings of (Gimenez et al., 2008) and (Karl and Werner, 2010), see below). In contrast, both right MTL NAA/Cho and NAA/Cr are each positively associated with body mass, suggesting an association between right hippocampal NAA and body mass without commensurate Cho or Cr increases. Absolute quantification of NAA is clearly warranted to validate these findings. Nevertheless, NAA deficits were evident despite the observation that hippocampal volume assessed morphometrically did not differ between VFD and LFD subjects (Mathew et al., 2003). The data suggest low left MTL NAA not detected in our previous report (Mathew et al., 2003), where right and left MTL neurometabolite ratios were collapsed, since mean neurometabolite values within each group did not differ by laterality. In that analysis, mean MTL NAA/Cr did not differ between groups, while mean MTL Cho/Cr ratios were significantly higher in VFD subjects. It is of note that low left, but not right, hippocampus NAA is similar to the low NAA observed in VFD ACC. However, both hippocampi differed from ACC in the VFD condition by manifesting high right and left Cho/Cr, in contrast to the absence of choline effects in ACC.
With relevance to early life stress, two conditions are of interest: perinatal stress and post traumatic stress disorder. Poland et al. (1999) reported reduced NAA in the left prefrontal cortex in adult rats subjected to perinatal stress. Moreover, given that neonatal intensive care exposes preterm neonates to a series of procedures resulting in prolonged stress (Anand, 2000), preterm birth can be considered a form of early life stress. Using MRSI and the linear combination model-fitting (LCModel) for absolute metabolite quantification, Gimenez et al. (2008) showed significant decrease in absolute NAA and Cr in the left MTL in adolescents with a history of preterm birth. However, they found no significant differences in absolute Cho, though the NAA/Cho ratio was significantly reduced in the left MTL (Gimenez et al., 2008). Alternatively, a recent meta-analysis of MRSI of the hippocampus in PTSD patients identified 12 studies with two types of control groups: healthy controls and traumatized non-PTSD controls. The meta-analysis that compared the PTSD group to the healthy control group revealed a significant decrease in the absolute NAA and an increase in the Cho/Cr ratio only in the left hippocampus of PTSD patients (Karl and Werner, 2010). The above mentioned findings illustrate the importance and complexity of hippocampal neurometabolites, while suggesting a correlation between perinatal/PTSD stressors, lower NAA and Cr, and higher Cho in the left hippocampus. Simultaneously, the findings of the PTSD meta-analysis suggest asymmetrical hippocampal neurometabolite changes consistent with the findings of the current study.
Although previous work has consistently shown neurohormonal differences between VFD’s and controls at various age points (Coplan et al., 1998; Coplan et al., 2001; Mathew et al., 2002; Smith et al., 2001), the neural substrate elaborating neurochemical differences in VFD, may be asymmetrically represented, and may model those differences observed in specific psychiatric entities. Thus, measures based on lumbar or cisternal CSF analyses represent a pooling of contributions of right and left sided structures. In the current report, there is a suggestion for specialized functions vis-à-vis left versus right hippocampus in relationship to adverse early rearing, body mass and neurometabolite inter-relationships. Several lines of preclinical and clinical evidence are concordant with the view that affective vulnerability is associated with compromise of left hippocampal neuronal viability. However, we are unaware of data positively linking relatively high right hippocampal NAA with body mass.
In rodents, neonatal novelty exposure modulates hippocampal volumetric asymmetry (Caldji et al., 2000). Early environmental manipulation affected hippocampal learning, hippocampal volume and cerebral lateralization, with early adversity corresponding to right volumetric dominance. Right functional neuronal dominance in the MTL region has also been observed in patients with schizophrenia (Qiu et al., 2009; Wang et al., 2001), panic disorder (for review, see (Wiedemann et al., 1999)), major depressive disorder (Bremner et al., 2000; Kronmuller et al., 2009) and in unsuccessful psychopaths (Raine et al., 2004). First episodes of major depression are associated with aberrant right>left volumetric asymmetry of hippocampus, suggesting lateralized abnormalities of neuronal viability either before or during an index depressive episode (Frodl et al., 2002). Kawakami et al. (2003) state that “despite implications for higher order functions of the brain, little is currently known about the molecular basis of left-right asymmetry of the brain”. The NMDA receptor plays an important role in a range of brain functions. Of the four identified murine NMDA-associated epsilon subunits, the authors report asymmetrical allocation only of the NMDA receptor epsilon 2 (NR2) subunits in rodent homogenates of right and left hippocampal formation. Extrapolation of these studies to other species suggests that the asymmetrical allocation of these subunits may result in differential ability to express hippocampal synaptic plasticity.
Caveats include the use of ratios rather than absolute quantification of neurometabolite ratios. In the current study, however, there is combined statistical evidence of low left NAA, irrespective of denominator used, in VFD subjects versus normally-reared control subjects. Second, the number of bilaterally represented hippocampi is relatively low due to shimming difficulties. Nevertheless, significant effects are noted when separately analyzing the data by cerebral hemisphere. Third, there was no tissue segmentation performed, and the possibility exists that differential gray/white matter proportions in voxels between right and left side could have contributed to reported differences in NAA/Cho. In summary, early life stress in the nonhuman primate form of VFD rearing in comparison to normally reared control subjects is associated with low left NAA concentrations within the left hippocampus. This effect remains significant when controlling for confounds introduced by body mass, and choline and creatine neurometabolite ratios. Given that NAA reduction in the left MTL may reflect low neuronal viability of the hippocampus, validation of this hypothesis is recommended, potentially using new ways of neural stem and progenitor cells identification developed by Manganas et al. (2007). In addition, follow-up neurohistological examinations for histopathological correlates of low hippocampal NAA, in the absence of gross volume changes, appear warranted.
The subjects for all nonhuman primate MRSI scans reported herein represent a re-analysis of a sample of nineteen adult male bonnet macaques (Macaca radiata) that had been reared in infancy under VFD (N = 10) or control (N = 9) conditions (Mathew et al., 2003). Mean age at the time of scanning for VFD subjects was approximately 9 years of age (see Table 1), corresponding to early to full adulthood. Nine of ten VFD-reared subjects were matched to controls by age and weight. Rearing groups did not differ in either mass or age (see Table 1). Rearing groups, as previously reported, also did not differ in terms of housing conditions or amount of anesthetic received. All procedures were conducted according to protocol approved by the SUNY-Downstate IACUC.
Briefly, mother-infant dyads were group-housed in pens of 5-7 dyads each and stabilized for at least four weeks prior to VFD onset. After infants reached at least 2 months of age, dyads were subjected to a standard VFD procedure that involved eight alternating 2-week blocks in which maternal food was either readily accessible (low foraging demand) or more difficult to obtain (high foraging demand).
Difficulty in obtaining food for the mothers was manipulated through the use of a foraging cart, a device in which food rations can be hidden in wood chips. No caloric restriction was present in the VFD procedure, and normal maternal and infant weights were maintained (Rosenblum and Paully, 1984). Following the VFD procedures, offspring were first housed with their mothers and then in standard peer social groups once they reached the juvenile phase of development, with no subsequent experimental manipulations that could confound the VFD-rearing effects.
The MRSI data were obtained using a GE 1.5 T instrument at Columbia University. Subjects were scanned under anesthesia induced by intramuscular administration of a combination of alphaxalone and alphadolone (Saffan; for full details of anesthesia and pre-scanning preparation, see (Mathew et al., 2003)). Following standard T1-weighted sagittal brain magnetic resonance imaging (MRI) scout images, a 2-section T1-weighted axial/oblique MRS localizer image was acquired at the same slice locations, using the same slice thickness (15 mm), spacing (2.5 mm), and angulation (parallel to the Sylvian fissure) as in the MRI scout scan. Next, a multislice proton MRS image scan was performed according to methods described by Duyn et al. (1993). Single voxels were placed bilaterally within the medial temporal lobe (for voxel placement, see (Mathew et al., 2003)), and aligned to represent hippocampus. In some instances, poor quality spectra due to shimming precluded analysis of one or both hippocampi. Long echo time metabolites were ascertained in vivo, using software specifically developed by Shungu et al. (1992).
were performed by two investigators blinded to the subjects’ rearing group status and using an MRSI analysis procedure and software package (XsOsNMR) as previously described (Shungu et al., 1992).
Because of technical difficulties due to shimming, only 17 scans were available for right MTL neurometabolite ratios and 13 scans for left MTL. Eleven subjects had MTL spectral values available bilaterally. Of note, there was no significant laterality effect for positive MTL spectral acquisition rates (Mathew et al., 2003). In the current analysis, left and right MTL neurometabolite ratios, rather than being collapsed across hemisphere, were analyzed as discrete bilateral regions of interest (ROIs). Moreover, as an extension of the Mathew et al. (2003) study, NAA/Cho was included as a neurometabolite ratio. The a priori goal of the MRSI studies was to detect reductions of MTL NAA levels, as a possible marker of hippocampal neuronal viability. To clarify findings based on ratios, we performed additional statistical regression analyses. A correlation matrix of all study variables was performed to ascertain the relationship among adverse early rearing, adult body mass and age, and neurometabolite ratios.
Significance was set at p ≤ .05, two-tailed. Bonferroni correction controlling for multiple comparisons of the four MTL NAA variables [two groups (VFD vs. LFD) x two NAA ratios (NAA/Cr & NAA/Cho)] -- p ≤ .0125 was required for significance for rearing group NAA differences. Bonferroni correction was not used for the regression analyses or correlation matrix.
Research supported by: National Institute of Mental Health Grant MH 59990 and National Alliance for Research on Schizophrenia and Depression Mid-Investigator Award (JDC); National Alliance for Research on Schizophrenia and Depression Young Investigator’s Award (SJM), Psychiatric Institute Research Support Grant and by National Institute of Mental Health research fellowship T32-MH15144 (SJM).
The authors thank Bruce Scharf, Douglas Rosenblum, Mohammad Arif, Shirn Baptiste, and Manuel de la Nuez for their invaluable contributions to this project.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.