The main goal of this study was to examine the long-term consequences of infant maltreatment on brain WM tracts of adolescent rhesus monkeys and to determine whether they were related to the elevated cortisol levels reported in these maltreated animals during infancy [8, 9]. We also examined whether alterations in brain WM microstructure were related to the increased aggressive behavior previously reported in the maltreated animals during adolescence . To do this we used measures of microstructural integrity, specifically FA, RD, and AD, calculated from DTI scans. We chose this technique because of its sensitivity to changes in WM microstructure, such as myelin thickness and axon/microtubule density [39–41]. These are neuronal characteristics that can affect the timing and speed of intercellular communication [52, 53], and can therefore affect behavior . FA increases significantly in brain WM tracts throughout primate development, and is accompanied by decreases in RD and few changes in AD [28, 29, 34–36]. These developmental changes in measures of axon microstructure suggest a global increase in tract integrity mainly due to increases in myelin from childhood to adulthood. Brain region-specific increases in FA are also observed after training on visuo-motor tasks  and with acquirement of new cognitive skills, such as reading and math, in parallel to decreases in RD, but no changes in AD [47, 48, 53, 56]. This suggests that these experience-related and region-specific increases in FA are due to increases in myelin and underlie behavioral and cognitive improvements. In contrast, reduced FA, associated in most regions with elevated cortisol during infancy and with increased concurrent aggression in one of the clusters, was detected here in adolescent rhesus monkeys that experienced infant maltreatment. Our findings are consistent with previous reports in human individuals that experienced childhood maltreatment [25, 27] or other forms of early life stress [23, 24] and in other nonhuman primate models of adversity , as well as in several mood and anxiety disorders [66, 95], with significant overlap with the regions affected in the current study.
To our knowledge, this is the first DTI study to examine the long-term effects of infant maltreatment on brain WM tract integrity in a nonhuman primate model. It is also the first to examine the associations of brain structural alterations with infant cortisol elevations and concurrent social behavior. Our findings show alterations in brain WM tract integrity measured using DTI in adolescent rhesus monkeys with histories of infant maltreatment. Decreased WM integrity (that is, FA) was found in maltreated subjects in the CC, occipital WM, EML, and brainstem, in comparison to controls. These regional FA decreases were paralleled by increases in RD, but no changes in AD, suggesting that the alterations in tract microstructural integrity in these brain regions were likely due to reduced myelin [42, 45–48, 93, 94]. An exception was the brainstem cluster, where no RD differences were found between groups. Basal plasma cortisol levels measured when the individuals were one month old, when abuse rates were highest , were negatively correlated with FA in all regions except for the brainstem cluster. This suggests that maltreatment at that early age caused stress-induced elevations in cortisol that could have potentially contributed to the long-term brain WM alterations reported. However, future studies are needed to examine causality in this relationship.
One of the clusters with lower FA in maltreated animals than controls was located in the lateral aspect of the medial midbody of the CC . The CC is the largest WM tract in the brain conveying interhemispheric fibers important for integration of information between cortical regions in both hemispheres . Because these fibers are some of the last to myelinate [31, 32, 36, 37], finding alterations in the CC is consistent with the view that areas undergoing active myelination or other protracted developmental processes are especially vulnerable to environmental experience [97, 98]. Alterations in the CC have also been reported in several studies of maltreated children, with reduced CC volume reported in maltreated children [99, 100], a difference that appears to be related to a failure to show the typical age-related increase in volume . Reduced CC size has also been reported in adults with histories of childhood maltreatment , suggesting that these CC alterations are persistent. Decreased FA in the CC of maltreated children  and adults who have experienced various forms of early life stress  has also been reported. The findings of the current study are also consistent with findings of reduced CC size in other nonhuman primate models of adverse early experience . Our findings of reduced WM integrity in the CC medial midbody region, which carries some prefrontal but mostly frontal motor and somatosensory fibers , could result in group differences in integration of motor and somatosensory information. The reduced interhemispheric integration reported here and in human studies of childhood maltreatment could contribute to behavioral alterations and psychopathology, an idea supported by similar CC alterations reported in anxiety and mood disorders .
The location of the three clusters identified in occipital WM suggest that the tracts affected could include short intra-occipital fiber systems (possibly part of the forceps major, an interhemispheric tract that connects occipital cortices in both hemispheres), and/or the caudal portion of the ILF, a long cortico-cortical association tract that courses through occipital, parietal, and temporal cortices . However, this can’t be corroborated without running additional tractography analyses. Interestingly, reduced FA has been reported in the forceps major of adolescents with histories of child maltreatment  and in the caudal portion of the ILF in adolescents that witnessed domestic violence as children . The ILF is part of the ventral visual pathway which is important for object identification , face processing , and emotional memory [110, 111]. Along these lines, alterations in WM microstructure of the ILF have been observed in several mood and anxiety disorders. For example, decreased FA in the ILF at the level of the occipital lobe has also been found in patients with depression [112, 113] and bipolar disorder [114, 115]. Thus, it is possible that decreases in microstructural integrity of occipital WM, likely involving the ILF, could affect visual and face processing, as well as emotion/mood processes.
The negative correlation of FA with aggressive behavior detected in occipital WM is difficult to explain. Most neuroimaging studies involving neural substrates of aggression implicate structural and/or functional abnormalities in frontal brain circuits [116, 117],although many of these studies have been done in patients with schizophrenia. Decreased FA in the anterior commissure (AC) has also been reported in violent youth with bipolar disorder, and FA in the AC was negatively correlated with aggression . However, this study was done in a clinical population making it difficult to integrate with the findings reported here. Increased occipital WM volume has been reported in adult violent offenders , but to our knowledge no other occipital alterations have been associated with aggression. Interestingly, a recent study comparing neural systems supporting social cognition in chimpanzees and bonobos reported that chimpanzees (known to be more aggressive than bonobos) had higher FA in occipital WM and bigger occipital GM volumes than bonobos , suggesting a potential association between aggression and FA in occipital WM in these species. The discrepancy of the directionality of the correlation with our findings could be explained by factors such as species-specific differences in neural substrates of aggression, or age at measurement. Given the paucity of research on the neural substrates of aggression, particularly in children, the interpretation of our findings is difficult. The visual cortices located near the cluster in which FA and aggression were correlated are part of attentional networks , and thus alterations in these circuits could reflect more general alterations in attention that might be better reflected by other behaviors not measured in the current study. It also has to be noted that our small sample size is a limitation for these studies, which might have been underpowered to detect other significant associations.
The WM cluster located lateral to the pulvinar thalamic nucleus and dorsal to the hippocampus seems to be the EML based on rhesus brain atlases . The EML contains both thalamo-cortical and cortico-thalamic fibers connecting the thalamus with parietal, temporal, occipital, cingulate, motor and PFC . Although without performing tractography it is difficult to precisely identify the specific thalamic nuclei and cortical regions connected by the affected tracts, based on the rostro-caudal location of this cluster the fibers affected likely connect the thalamus with occipital or temporal cortices . Interestingly, thalamo-cortical systems modulate amygdala activity, and are involved in the perception of fear . Cortico-thalamic circuits are implicated in the pathogenesis of mood disorders . Thus, our findings of reduced structural integrity in EML suggest potential alterations in cortico-thalamic and thalamo-cortical circuits that could contribute to deficits in emotional regulation reported above in maltreated animals.
The brainstem cluster where FA was lower in maltreated animals than controls was difficult to identify anatomically due to the low MRI contrast in this region. However, as described above, its location matches the position of the CTT . The CTT is a pathway containing descending fibers from midbrain nuclei that project to the olivary complex, as well as ascending fibers originating in the pontine and medullary reticular formation that project to the thalamus . These are brainstem pathways that carry and coordinate somatosensory and somatomotor information. MRI studies report lesions in the CTT in neurodegenerative and neurodevelopmental disorders, linked with motor and cognitive deficits . This was the only region where group differences in FA (lower in maltreated subjects than in controls) were not related to the increased levels of cortisol during infancy in the maltreated animals, suggesting that the effects of maltreatment on this WM could be associated with other aspects of the early experience.
There are limitations to the DTI method as applied here. Most are due to the low spatial resolution of the diffusion data acquired in the relatively small rhesus brain. At this resolution partial voluming effects can make interpreting or finding results difficult. The TBSS analysis applied here addresses this limitation by using only voxels from the centers of large WM tracts in individual subjects. Partial voluming can also make registration difficult, which is another reason why we used the nonlinear registration built into the TBSS processing pipeline to perform our voxelwise analyses. The low angular resolution (that is, the small number of directions acquired for the DTI data), especially when combined with the low spatial resolution of our data, also makes accurate probabilistic tractography difficult, which is why it was not performed in these studies. Tractography would be helpful in future studies to determine the exact tracts affected in the clusters with group differences, although it would not help in determining the directionality of the affected fibers.
The correlations between infant cortisol and WM integrity found in the current study suggest that early life stress has long-term effects on brain WM in regions previously reported as vulnerable to childhood maltreatment in humans, and that are also altered in anxiety and mood disorders. One possible mechanism could be through the effects of elevated levels of glucocorticoids (GCs), in this case cortisol, on the development of WM . Oligodendrocytes that form the myelin sheath express both intracellular glucocorticoid and mineralocorticoid receptors , and recent evidence suggests that GCs suppress proliferation of oligodendrocyte precursor cells in GM and WM . Developmental studies also provide evidence that GCs modulate oligodendrocyte differentiation and myelogenesis via regulation of key oligodendroglial proteins such as myelin basic protein (MBP) , and that the effects of synthetic GCs differ as function of gestational age, with decreases in MBP immunoreactivity and numbers of oligodendrocytes associated with younger ages of GC exposure . Taken together, these studies suggest that myelination is sensitive to GCs during development, making it possible for early life stress, via elevated cortisol levels, to affect brain WM development. The associations detected in our studies between decreased FA and basal cortisol levels at one month are consistent with this possibility, although the causality of this relationship needs to be tested in future studies. Due to the strong role of brain WM in behavioral control, for example, , GC-induced alterations in brain WM development could potentially lead to the alterations reported in maltreated monkeys, including increased aggression. Our findings also open new questions and hypotheses that need to be empirically tested. Does maltreatment lead to altered function of the affected circuits? When do these differences emerge and how they unfold? Prospective, longitudinal studies beginning at birth are necessary to address these important developmental questions in the context of maltreatment to determine the most beneficial timing and type of potential treatments, as well as intervention and prevention strategies.