Introduction
Early psychosocial adversities have been associated with poorer emotional and physical functioning, with higher vulnerability to further trauma exposure, and with higher rates of adult psychopathology (Felitti et al. Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards, Koss and Marks1998; Rutter, Reference Rutter2002; Edwards et al. Reference Edwards, Holden, Felitti and Anda2003; Moffitt et al. Reference Moffitt, Caspi and Rutter2005; Anda et al. Reference Anda, Felitti, Bremner, Walker, Whitfield, Perry, Dube and Giles2006; Teicher et al. Reference Teicher, Samson, Sheu, Polcari and McGreenery2010; Wermter et al. Reference Wermter, Laucht, Schimmelmann, Banaschweski, Sonuga-Barke, Rietschel and Becker2010). Bipolar disorder (BD) is a progressive and accelerating condition with a cyclical alternating pattern of manic and depressive episodes, and has been associated with neurostructural changes, cognitive deterioration and accumulated medical co-morbidity. A history of childhood maltreatment is highly prevalent in patients with BD, and has been associated with an earlier onset of the disorder, worse clinical course, and higher lifetime suicidal ideation and suicide attempts (McIntyre et al. Reference McIntyre, Soczynska, Mancini, Lam, Woldeyohannes, Moon, Konarski and Kennedy2008; Daruy-Filho et al. Reference Daruy-Filho, Brietzke, Lafer and Grassi-Oliveira2011; Brietzke et al. Reference Brietzke, Mansur, Soczynska, Kapczinski, Bressan and McIntyre2012).
An impressive series of studies in healthy subjects associated adverse childhood experiences (ACE) with in vivo measures of adult white matter (WM) microstructure. Given the microscopic structure of WM, in normal conditions the integrity of myelinated axons limits the diffusion of water in directions other than along the main axis of the fibre. This tendency to diffuse in one direction as opposed to all others, termed anisotropy, reflects the integrity of axons and myelin sheaths and the bundle coherence of WM tracts, and can be estimated in vivo through the application of diffusion-tensor imaging (DTI) techniques (Basser et al. Reference Basser, Mattiello and LeBihan1994; Le Bihan, Reference Le Bihan2003; Taylor et al. Reference Taylor, Hsu, Krishnan and MacFall2004b ). These allow estimation of the tendency to diffuse along the principal direction of the fibre (axial diffusivity; AD) or perpendicular to axonal walls (radial diffusivity; RD) (Song et al. Reference Song, Sun, Ramsbottom, Chang, Russell and Cross2002), and to calculate the variance of the direct measures of the diffusion magnitude in these directions (fractional anisotropy; FA).
Early severe socio-emotional deprivation in Eastern European orphanages has been found to be associated with reduced FA in the left uncinate fasciculus (Eluvathingal et al. Reference Eluvathingal, Chugani, Behen, Juhasz, Muzik, Maqbool, Chugani and Makki2006), paralleling glucose hypometabolism in limbic and paralimbic structures (Chugani et al. Reference Chugani, Behen, Muzik, Juhasz, Nagy and Chugani2001). A pivotal study in healthy adults associated parental verbal abuse during childhood with a significantly reduced FA in the arcuate fasciculus in the left superior temporal gyrus, in the cingulum bundle by the posterior tail of the left hippocampus, and in the left body of the fornix (Choi et al. Reference Choi, Jeong, Rohan, Polcari and Teicher2009). The same group then documented an increased mean diffusivity (MD) and RD with decreased FA in the corpus callosum and corona radiata of adults exposed to peer verbal abuse (Teicher et al. Reference Teicher, Samson, Sheu, Polcari and McGreenery2010), and a reduction of FA values in the inferior longitudinal fasciculus of the left lateral occipital lobe in adults who had witnessed domestic violence during childhood (Choi et al. Reference Choi, Jeong, Polcari, Rohan and Teicher2012). These data suggested an association between childhood maltreatment and the development of fibre pathways that convey the adverse experience to frontal, temporal or limbic regions (Teicher et al. Reference Teicher, Tomoda and Andersen2006). A recent study confirmed that adolescents exposed to childhood maltreatment had lower FA in the superior longitudinal fasciculi, right cingulum bundle projecting to the hippocampus, left inferior fronto-occipital fasciculus, and splenium of the corpus callosum, with lower values being associated with emerging depression during follow-up (Huang et al. Reference Huang, Gundapuneedi and Rao2012), thus supporting the hypothesis that WM disruption in circuitries critical for emotional and cognitive processing could be linked with mood disorder psychopathology (Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ). However, the effects of stress on WM are not limited to ACE, because post-exposure combined reduction in hippocampal volume and connectivity with the prefrontal cortex have been found to mark a maladaptive response to stressful military service (Admon et al. Reference Admon, Leykin, Lubin, Engert, Andrews, Pruessner and Hendler2013).
Changes in WM microstructure have indeed been proposed as structural biomarkers of BD, independent of the effects of ongoing drug treatments. Widespread changes in DTI measures of WM integrity have been reported in untreated BD depressed patients (Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ), in drug-naive BD patients with mania (Adler et al. Reference Adler, Adams, DelBello, Holland, Schmithorst, Levine, Jarvis and Strakowski2006), in untreated BD patients with first-episode psychosis (Lu et al. Reference Lu, Zhou, Keedy, Reilly and Sweeney2011) and in high-risk unaffected relatives of patients with BD (Sprooten et al. Reference Sprooten, Sussmann, Clugston, Peel, McKirdy, Moorhead, Anderson, Shand, Giles, Bastin, Hall, Johnstone, Lawrie and McIntosh2011). These studies consistently documented an increased diffusivity with reduced FA during illness phases of both manic and depressive polarity, while the few observations in stable euthymic patients showed increased FA in patients with BD compared with controls (Houenou et al. Reference Houenou, Wessa, Douaud, Leboyer, Chanraud, Perrin, Poupon, Martinot and Paillere-Martinot2007; Wessa et al. Reference Wessa, Houenou, Leboyer, Chanraud, Poupon, Martinot and Paillere-Martinot2009). Long-term administration of lithium, the mainstay for the treatment of BD, has been associated with higher diffusivity along the principal fibre axis (AD, λ 1), thus possibly counteracting the detrimental changes in WM microstructure associated with BD (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). These findings suggest that WM diffusivity changes might be of clinical relevance in BD, and that DTI could provide new biomarkers to estimate susceptibility to the disorder, to track its lifetime progression, to identify new targets for treatment, and to predict and monitor treatment efficacy (Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c , Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013).
The documented associations between ACE and WM microstructure in healthy subjects, between BD and WM microstructure, and between ACE and outcome of BD, led us to hypothesize that ACE could be associated with altered DTI measures of WM integrity in patients with BD. In the present study, using tract-based spatial statistics (TBSS), we tested this hypothesis in a homogeneous sample of patients affected by a major depressive episode in the course of BD.
Method
Participants and clinical assessment
We studied 80 consecutively admitted in-patients affected by a major depressive episode without psychotic features, with a diagnosis of BP type I (Structured Clinical Interview for DSM Disorders). Patients were either drug-free (n = 54) or treated with lithium (n = 26). Exclusion criteria were: additional diagnoses on Axis I, mental retardation on Axis II, pregnancy, major medical and neurological disorders, history of drug or alcohol abuse or dependency. Physical examination, laboratory tests and electrocardiograms were performed at admission. No patient had received electroconvulsive therapy within 6 months prior to study enrolment. After complete description of the study to the participants, written informed consent was obtained. The study was approved by the local ethical committee.
Severity of ACE was rated on the Risky Families Questionnaire (RFQ) (Taylor et al. Reference Taylor, Eisenberger, Saxbe, Lehman and Lieberman2006) after functional magnetic resonance imaging (fMRI) scanning. The RFQ has been adapted from an instrument originally developed to assess the relationship of family stress to mental and physical health outcomes in adulthood (Felitti et al. Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards, Koss and Marks1998). The instrument is aimed at rating the degree of harsh parenting with overt family conflict and deficient nurturing experienced by the children in their familial environment. Previous research has validated this questionnaire against clinical interviews conducted and coded by trained clinical interviewers; the dual assessments (questionnaire and interview) demonstrated high agreement and reliability (Taylor et al. Reference Taylor, Lerner, Sage, Lehman and Seeman2004a ). This approach has been proven successful in detecting the grey matter structural and functional brain correlates of ACE in adult life (Taylor et al. Reference Taylor, Eisenberger, Saxbe, Lehman and Lieberman2006; Benedetti et al. Reference Benedetti, Radaelli, Poletti, Falini, Cavallaro, Dallaspezia, Riccaboni, Scotti and Smeraldi2011b ).
Image acquisition
DTI was performed on a 3.0 Tesla scanner (Gyroscan Intera; Philips, The Netherlands) using spin-echo echo-planar imaging and the following parameters: repetition time (TR) = 8753.89 ms; echo time (TE) = 58 ms; field of view 231.43 mm (AP), 126.50 mm (FH), 240.00 mm (RL); acquisition matrix 2.14 × 2.71 × 2.31; 55 contiguous, 2.3-mm thick axial slices reconstructed with in-plane pixel size 1.88 × 1.87 mm; sensitivity encoding acceleration factor = 2; one b0 and 35 non-collinear directions of the diffusion gradients; b value = 900 s/mm2. Fat saturation was performed to avoid chemical shift artifacts. On the same occasion and using the same magnet, 22 turbo spin echo, T2 axial slices (TR = 3000 ms; TE = 85 ms; flip angle = 90°; turbo factor 15; 5-mm-thick, axial slices with a 512 × 512 matrix and a 230 × 230 mm2 field of view) were acquired to rule out brain lesions.
Data processing and analyses
Image analyses and tensor calculations were carried out using the ‘Oxford Center for Functional Magnetic Resonance Imaging of the Brain Statistical Library’ (FSL 4.1.4; www.fmrib.ox.ac.uk/fsl/index.html) (Smith et al. Reference Smith, Jenkinson, Woolrich, Beckmann, Behrens, Johansen-Berg, Bannister, De Luca, Drobnjak, Flitney, Niazy, Saunders, Vickers, Zhang, De Stefano, Brady and Matthews2004; Woolrich et al. Reference Woolrich, Jbabdi, Patenaude, Chappell, Makni, Behrens, Beckmann, Jenkinson and Smith2009). First, each of the 35 DTI volumes was affine registered to the T2-weighted b = 0 volume using FLIRT (FMRIB's Linear Image Registration Tool) (Jenkinson & Smith, Reference Jenkinson and Smith2001). This corrected for motion between scans and residual eddy-current distortions present in the diffusion-weighted images. In addition, trained researchers blind to diagnosis manually inspected each volume of each image to check for head motion artifacts: scans rated as 0 (none) had little or no detectable motion artifact, those rated as 1 (mild) had enough detectable motion to result in subtle concentric bands, 2 (moderate) had significant banding while those rated as 3 (severe) were so extreme that the data were deemed unreliable for analyses (Blumenthal et al. Reference Blumenthal, Zijdenbos, Molloy and Giedd2002). Anisotropy can be estimated through the application of diffusion-sensitizing gradients and the calculation of elements of the diffusion tensor matrix, i.e. the three eigenvalues λ 1, λ 2 and λ 3 (Basser et al. Reference Basser, Mattiello and LeBihan1994; Le Bihan, Reference Le Bihan2003; Taylor et al. Reference Taylor, Hsu, Krishnan and MacFall2004b ). The tendency to diffuse along the principal direction of the fibre (AD, λ 1) reflects the integrity of axons and myelin sheaths, and the bundle coherence of WM tracts (Boretius et al. Reference Boretius, Escher, Dallenga, Wrzos, Tammer, Bruck, Nessler, Frahm and Stadelmann2012). An increase in RD (the average of λ 2 and λ 3), perpendicular to axonal walls, suggests disrupted myelination (Song et al. Reference Song, Sun, Ramsbottom, Chang, Russell and Cross2002). MD (average of λ 1, λ 2 and λ 3) is a measure of the average molecular motion, independent of tissue directionality. FA is the square root of the sum of squares (SRSS) of the diffusivity differences, divided by the SRSS of the three diffusivities. After removal of non-brain tissue (Smith, Reference Smith2002), least-square fits were performed to estimate the FA, eigenvector, and eigenvalue maps. MD was defined as the mean of all three eigenvalues [(λ 1 + λ 2 + λ 3)/3], AD as the principal diffusion eigenvalue (λ 1), and RD as the mean of the second and third eigenvalues [(λ 2 + λ 3)/2].
Next, all individuals’ volumes were skeletonized and transformed into a common space as used in TBSS (Smith et al. Reference Smith, Jenkinson, Johansen-Berg, Rueckert, Nichols, Mackay, Watkins, Ciccarelli, Cader, Matthews and Behrens2006, Reference Smith, Johansen-Berg, Jenkinson, Rueckert, Nichols, Miller, Robson, Jones, Klein, Bartsch and Behrens2007). TBSS focuses on the centres of all fibre bundles that are common to the participants (the most compact WM skeleton), thus improving the probability that the given spatial voxels contain data from the same part of the same WM tract of each participant. Briefly, all volumes were non-linearly warped to the FMRIB58_FA template supplied with FSL (http://www.fmrib.ox.ac.uk/fsl/tbss/FMRIB58_FA.html) and normalized to the Montreal Neurological Institute (MNI) space, by use of local deformation procedures performed by FMRIB's Non-Linear Image Registration Tool (FNIRT) (www.fmrib.ox.ac.uk/fsl/fnirt/index.html), a non-linear registration toolkit using a b-spline representation of the registration warp field (Rueckert et al. Reference Rueckert, Sonoda, Hayes, Hill, Leach and Hawkes1999). The common template used in the present study is a high-resolution average of 58 FA volumes from healthy male and female subjects aged 20–50 years. All warped FA volumes were visually inspected for accuracy, which is especially pertinent when analysing datasets with broad age ranges with relatively large interindividual variability in brain size and architecture. FNIRT has been shown to perform native-to-standard warping adequately across several age groups, including children and adolescents (Westlye et al. Reference Westlye, Walhovd, Dale, Bjornerud, Due-Tonnessen, Engvig, Grydeland, Tamnes, Ostby and Fjell2010). Next, a mean FA volume of all subjects was generated and thinned to create a mean FA skeleton representing the centres of all common tracts. We thresholded and binarized the mean skeleton at FA > 0.20 to reduce the likelihood of partial voluming in the borders between tissue classes, yielding a mask of 137 833 WM voxels. Individual FA values were warped onto this mean skeleton mask by searching perpendicular from the skeleton for maximum FA values. Using maximum FA values from the centres of the tracts further minimizes confounding effects attributable to partial voluming (Smith et al. Reference Smith, Jenkinson, Johansen-Berg, Rueckert, Nichols, Mackay, Watkins, Ciccarelli, Cader, Matthews and Behrens2006). The resulting tract invariant skeletons for each participant were fed into voxelwise permutation-based cross-subject statistics. Similar warping and analyses were used on MD, AD and RD data sampled from voxels with FA > 0.20.
Voxelwise DTI analyses were performed using non-parametric permutation-based testing (Nichols & Holmes, Reference Nichols and Holmes2002) as implemented in randomise in FSL. We tested for linear effects of ACE on FA, MD, AD and RD across the WM skeleton with general linear models. We accounted for the effects of nuisance covariates that could influence WM structure: age (Kochunov et al. Reference Kochunov, Thompson, Lancaster, Bartzokis, Smith, Coyle, Royall, Laird and Fox2007), sex (Herting et al. Reference Herting, Maxwell, Irvine and Nagel2011), education (Foubert-Samier et al. Reference Foubert-Samier, Catheline, Amieva, Dilharreguy, Helmer, Allard and Dartigues2012), and duration of lithium treatment in months (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). Threshold-free cluster enhancement (TFCE) (Smith & Nichols, Reference Smith and Nichols2009) was used to avoid defining arbitrary cluster-forming thresholds and smoothing levels. TFCE is particularly useful when the spatial correlation length of the signal exceeds that of noise, as it is expected when studying WM tracts. It can be seen as a generalization of the cluster mass statistics (Bullmore et al. Reference Bullmore, Suckling, Overmeyer, Rabe-Hesketh, Taylor and Brammer1999), using spatial neighbourhood information in non-linear image processing to increase sensitivity and boosting the height of spatially distributed signals, without changing the location of their maxima. Voxelwise levels of significance, corrected for multiple comparisons, were then calculated with a standard permutation testing by building up the null distribution (across permutation of the input data) of the maximum (across voxels) TFCE scores, and then using the 95th percentile of the null distribution to threshold signals at corrected p < 0.05. The data were tested against an empirical null distribution generated by 5000 permutations for each contrast, thus providing statistical maps fully corrected for multiple comparisons across space. Corrected p < 0.05 in a minimum cluster size of k = 100 was considered significant.
In addition to FA and MD, we analysed the significance of the effects of ACE on the single eigenvalue λ 1 and RD. The eigenvalue λ 1 directly measures diffusion along the principal axis of the WM tracts (AD), thus accounting for those components of the diffusion anisotropy that are related to axonal structure, directionality and branching. RD measures diffusion perpendicular to the fibre, thus being directly influenced by the integrity of myelin sheaths. Previous work has shown that these measures are more able to capture the subtle WM abnormalities associated with bipolar illness and its treatment (Benedetti et al. Reference Benedetti, Absinta, Rocca, Radaelli, Poletti, Bernasconi, Dallaspezia, Pagani, Falini, Copetti, Colombo, Comi, Smeraldi and Filippi2011a , Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla c , Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013).
Results
Clinical and demographic characteristics of the sample are shown in Table 1.
Table 1. Clinical and demographic characteristics of the sample
Data are given as mean (standard deviation).
In agreement with earlier observations (Daruy-Filho et al. Reference Daruy-Filho, Brietzke, Lafer and Grassi-Oliveira2011), and confirming the clinical relevance of ACE in the present sample, higher ACE as rated on the RFQ were associated with a significantly earlier age of onset of illness (Pearson's r = 0.342, p = 0.002; see Fig. 1). The other measures were not associated with RFQ scores.
Fig. 1. Inverse correlation (––) between the severity of adverse childhood experiences measured using the Risky Families Questionnaire (RFQ) and onset of illness (Pearson's r = 0.342, p = 0.002). The worse the adverse childhood experiences, the earlier the onset of bipolar disorder. The dashed lines represent 95% confidence intervals.
ACE influenced DTI measures of WM microstructure. Patients with higher RFQ scores showed lower values of AD in several brain WM tracts (Fig. 2).
Fig. 2. White matter areas where axial diffusivity showed a significant inverse relationship with adverse childhood experiences (Risky Families Questionnaire scores). Voxels of significant difference are shown on the mean fractional anisotropy (FA) template of the studied sample. The colour bar refers to 1 – p values for the observed differences. Group differences are mapped onto the mean FA brain template. Numbers are z coordinates in the standard Montreal Neurological Institute (MNI) space. Images are in neurological convention (left side of the brain is represented on the right side of the pictures).
Differences in AD were observed in two main clusters, which included the corpus callosum and forceps minor, bilateral anterior and posterior cingulum bundle, bilateral corona radiata with anterior and posterior thalamic radiation, bilateral superior longitudinal fasciculus, left inferior fronto-occipital fasciculus and left uncinate fasciculus (see Table 2).
Table 2. Effect of ACE on axial diffusivity a
ACE, Adverse childhood experiences; MNI, Montreal Neurological Institute; R, right; L, left.
a In the first column, dimensions of clusters (number of voxels, mm3) and localization of signal peaks (MNI coordinates) are given for regions showing maximal differences of tract-based spatial statistics values (signal peak). The second column lists the white matter tracts significantly affected by ACE in the clusters.
No other effect was significant. In particular, no positive correlation was observed between ACE and AD in any region, and no effect of sex or clinical variables (severity of depression), or any effect of ACE on FA, RD and MD, survived the statistical threshold of p < 0.05 corrected for multiple comparisons across space.
Discussion
This is the first study to report an effect of ACE on WM integrity in bipolar patients. We observed that the severity of exposure to ACE was associated with decreased DTI measures of AD in many WM tracts connecting cortical and subcortical brain structures.
AD represents the water diffusivity parallel to the axonal fibres, reflecting the greater freedom of water to diffuse along the principal fibre axis rather than to travel across the surrounding myelin sheaths. Both myelin and axonal microstructure, including microtubules and neurofilaments (Kinoshita et al. Reference Kinoshita, Ohnishi, Kohshi and Yokota1999), contribute to this diffusion anisotropy. Studies on neurodevelopment have associated AD with fibre diameter and organization (Takahashi et al. Reference Takahashi, Ono, Harada, Maeda and Hackney2000). Over the human lifespan AD and MD decrease initially, and then increase later in life (Qiu et al. Reference Qiu, Tan, Zhou and Khong2008; Lebel et al. Reference Lebel, Gee, Camicioli, Wieler, Martin and Beaulieu2012), also reflecting changes in fibre coherence and tortuosity (Dubois et al. Reference Dubois, Dehaene-Lambertz, Perrin, Mangin, Cointepas, Duchesnay, Le Bihan and Hertz-Pannier2008). AD positively correlates with functional connectivity among brain regions, and with network-related aspects of human brain function also including, as behavioural correlates, the experience of emotions in response to stimuli (Baur et al. Reference Baur, Hanggi, Langer and Jancke2013) and higher-order abilities such as reasoning and cognitive flexibility (Borghesani et al. Reference Borghesani, Madhyastha, Aylward, Reiter, Swarny, Warner Schaie and Willis2013).
Here we observed that ACE were associated with a lower AD in key WM tracts contributing to the functional integrity of the brain. These included inter-hemispheric connections, limbic, and large frontal, parietal and fronto-occipital connections that have been proposed as key components of the brain network dysfunctions, putatively leading to the cognitive and emotional deficits typical of BD (Brambilla et al. Reference Brambilla, Bellani, Yeh and Soares2009; Benedetti et al. Reference Benedetti, Absinta, Rocca, Radaelli, Poletti, Bernasconi, Dallaspezia, Pagani, Falini, Copetti, Colombo, Comi, Smeraldi and Filippi2011a , Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla c ). ACE-associated alterations in WM structure could then contribute to the abnormal effective cortico-limbic connectivity observed with fMRI (Arnone et al. Reference Arnone, McIntosh, Chandra and Ebmeier2008; Rich et al. Reference Rich, Fromm, Berghorst, Dickstein, Brotman, Pine and Leibenluft2008), and to the affective and cognitive instability observed during the waxing and waning of illness episodes in BD. Examination of the available literature regarding the WM tracts where ACE are associated with decreased AD supports this hypothesis.
The corona radiata contains both descending and ascending axons that carry nearly all of the neural traffic from and to the cerebral cortex. The DTI measures of the corona radiata correlate with performance in executive function and information processing speed (Sasson et al. Reference Sasson, Doniger, Pasternak, Tarrasch and Assaf2012). Prefrontal cortical areas ensure the cognitive control of emotion experience and mood (Beauregard et al. Reference Beauregard, Lévesque and Bourgouin2001; Barrett et al. Reference Barrett, Mesquita, Ochsner and Gross2007). Their reactivity to stimuli parallels the mood-congruent biases in the processing of emotional stimuli during depression (Elliott et al. Reference Elliott, Rubinsztein, Sahakian and Dolan2002), and normalize after successful antidepressant treatment in BD (Benedetti et al. Reference Benedetti, Bernasconi, Blasi, Cadioli, Colombo, Falini, Lorenzi, Radaelli, Scotti and Smeraldi2007). The corona radiata showed a reduced integrity in patients with BD (Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ), soon at the beginning of illness (Barnea-Goraly et al. Reference Barnea-Goraly, Chang, Karchemskiy, Howe and Reiss2009; Pavuluri et al. Reference Pavuluri, Yang, Kamineni, Passarotti, Srinivasan, Harral, Sweeney and Zhou2009), and with lithium increasing AD proportionally to treatment duration (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). The corona radiata is sensitive to early stress in healthy humans, as documented by measures of reduced integrity in adults exposed to peer verbal abuse (Teicher et al. Reference Teicher, Samson, Sheu, Polcari and McGreenery2010).
The cingulum bundle is located within the WM of the cingulate gyrus and originates from the frontal lobe, rounds the corpus callosum from the genu extending to the splenium, travels along the ventral surface of the hippocampus, and terminates in the amygdala (Schmahmann et al. Reference Schmahmann, Pandya, Wang, Dai, D'Arceuil, de Crespigny and Wedeen2007). These connections are crucial for both motivational and emotional aspects of behaviour, integrating disparate modular brain systems involved in attention, memory and emotion regulation that have been considered at the core of mood dysregulation in BD (Davidson et al. Reference Davidson, Pizzagalli, Nitschke and Putnam2002). In mood disorders a higher directionality of water diffusion in the cingulum bundle is associated with better hedonic tone (Keedwell et al. Reference Keedwell, Chapman, Christiansen, Richardson, Evans and Jones2012), better cognitive performance (Schermuly et al. Reference Schermuly, Fellgiebel, Wagner, Yakushev, Stoeter, Schmitt, Knickenberg, Bleichner and Beutel2010) and less psychomotor slowing (Walther et al. Reference Walther, Hugli, Hofle, Federspiel, Horn, Bracht, Wiest, Strik and Muller2012), memory scores and hippocampal volume (Sexton et al. Reference Sexton, Mackay, Lonie, Bastin, Terriere, O'Carroll and Ebmeier2010). The cingulum is highly sensitive to early stress, with FA in this area being reduced in healthy humans exposed to parental verbal abuse and being inversely associated with ratings of depression, dissociation and limbic irritability (Choi et al. Reference Choi, Jeong, Rohan, Polcari and Teicher2009). Again, we had reported a breakdown in the architecture of these WM tracts in BD (Benedetti et al. Reference Benedetti, Absinta, Rocca, Radaelli, Poletti, Bernasconi, Dallaspezia, Pagani, Falini, Copetti, Colombo, Comi, Smeraldi and Filippi2011a ), and that lithium, contrary to stress, promotes a higher directionality of water diffusion in these tracts (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013).
The thalamic radiations are two-way fibre connections between the thalamus and cerebral cortex, forming a major part of the internal capsule and corona radiata (Liao et al. Reference Liao, Huang, Wu, Yang, Kuang, Du, Lui, Yue, Chan, Kemp and Gong2013). The anterior thalamic radiation (ATR) connects the anterior thalamic and dorsomedial nuclei with the prefrontal cortex and medial temporal cortex, and is part of the thalamo-fronto-striatal loops that have been implicated in the pathophysiology of both BD and schizophrenia (Buchsbaum et al. Reference Buchsbaum, Hazlett, Haznedar, Spiegel-Cohen and Wei1999; McIntosh et al. Reference McIntosh, Munoz Maniega, Lymer, McKirdy, Hall, Sussmann, Bastin, Clayden, Johnstone and Lawrie2008). Early life stress has been associated with significant reductions in FA in the anterior limb of the internal capsule in non-human primates (Coplan et al. Reference Coplan, Abdallah, Tang, Mathew, Martinez, Hof, Smith, Dwork, Perera, Pantol, Carpenter, Rosenblum, Shungu, Gelernter, Kaffman, Jackowski, Kaufman and Gorman2010). A reduced FA in the ATR has been consistently detected in patients with BD (McIntosh et al. Reference McIntosh, Munoz Maniega, Lymer, McKirdy, Hall, Sussmann, Bastin, Clayden, Johnstone and Lawrie2008; Sussmann et al. Reference Sussmann, Lymer, McKirdy, Moorhead, Munoz Maniega, Job, Hall, Bastin, Johnstone, Lawrie and McIntosh2009) soon after the beginning of illness (Pavuluri et al. Reference Pavuluri, Yang, Kamineni, Passarotti, Srinivasan, Harral, Sweeney and Zhou2009; Lu et al. Reference Lu, Zhou, Fitzgerald, Keedy, Reilly, Passarotti, Sweeney and Pavuluri2012), and in their unaffected relatives (McIntosh et al. Reference McIntosh, Job, Moorhead, Harrison, Lawrie and Johnstone2005), among whom FA in the internal capsule was inversely related to cyclothymic temperament (Sprooten et al. Reference Sprooten, Sussmann, Clugston, Peel, McKirdy, Moorhead, Anderson, Shand, Giles, Bastin, Hall, Johnstone, Lawrie and McIntosh2011), thus also suggesting a link with genetic vulnerability for the disorder. Converging ascending and descending medial forebrain bundle and ATR fibre tracts could mediate reward seeking and regulation of affective states in healthy and depressed humans (Schoene-Bake et al. Reference Schoene-Bake, Parpaley, Weber, Panksepp, Hurwitz and Coenen2010; Coenen et al. Reference Coenen, Panksepp, Hurwitz, Urbach and Madler2012). Executive function and processing speed were correlated with anisotropy of the ATR and uncinate fasciculus in late-life depression (Sexton et al. Reference Sexton, McDermott, Kalu, Herrmann, Bradley, Allan, Le Masurier, Mackay and Ebmeier2012). The posterior thalamic radiation (PTR) connects the pulvinar and lateral geniculate nuclei to the posterior parietal and occipital cortex, and its integrity correlates with behavioural reaction time in visuospatial attention tasks (Tuch et al. Reference Tuch, Salat, Wisco, Zaleta, Hevelone and Rosas2005). Impaired orientational coherence or reduced WM integrity has been consistently described in patients with major depressive disorder (MDD) (Liao et al. Reference Liao, Huang, Wu, Yang, Kuang, Du, Lui, Yue, Chan, Kemp and Gong2013) and with BD (Chan et al. Reference Chan, Yang, Chia, Woon, Lee, Keefe, Sitoh, Nowinski and Sim2010; Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ). Again, lithium counteracted these effects in the ATR and PTR by increasing AD (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013).
The corpus callosum provides the major commissural fibre bundles allowing interhemispheric communication. Corpus callosum integrity is crucial for the integration of sensory–motor functions, attention, language and memory, which are frequently impaired in bipolar patients (Brambilla et al. Reference Brambilla, Bellani, Yeh and Soares2009). Its structure is sensitive to early stress, because degree of exposure to peer verbal abuse correlates with increased RD and MD and decreased FA (Teicher et al. Reference Teicher, Samson, Sheu, Polcari and McGreenery2010). Abnormal myelination and morphometry have been consistently associated with BD (Bearden et al. Reference Bearden, van Erp, Dutton, Boyle, Madsen, Luders, Kieseppa, Tuulio-Henriksson, Huttunen, Partonen, Kaprio, Lonnqvist, Thompson and Cannon2011, Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ), and with dimensions of BD psychopathology such as aggression (Saxena et al. Reference Saxena, Tamm, Walley, Simmons, Rollins, Chia, Soares, Emslie, Fan and Huang2012) and suicide (Matsuo et al. Reference Matsuo, Nielsen, Nicoletti, Hatch, Monkul, Watanabe, Zunta-Soares, Nery and Soares2010). These abnormalities persist in euthymic conditions (Emsell et al. Reference Emsell, Leemans, Langan, Van Hecke, Barker, McCarthy, Jeurissen, Sijbers, Sunaert, Cannon and McDonald2013), when patients show signs of reduced fibre density (Torgerson et al. Reference Torgerson, Irimia, Leow, Bartzokis, Moody, Jennings, Alger, Van Horn and Altshuler2012), and again we found lithium to increase AD in the corpus callosum (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). Reduced FA correlates with the genetic liability risk for BD in unaffected relatives (Chaddock et al. Reference Chaddock, Barker, Marshall, Schulze, Hall, Fern, Walshe, Bramon, Chitnis, Murray and McDonald2009), and a progressive FA reduction during adolescence, contrary to the expected increase, has been observed in unaffected youth at high risk for BD and proposed as a vulnerability marker for future BD (Versace et al. Reference Versace, Ladouceur, Romero, Birmaher, Axelson, Kupfer and Phillips2010b ).
The superior longitudinal fasciculus is a huge association fibre tract connecting cortical areas of the frontal, parietal, temporal and occipital lobes, thus being involved in a wide range of functions including executive functioning and emotional regulation, in addition to language processing thanks to its connections to Broca and Wernicke areas (Wakana et al. Reference Wakana, Jiang, Nagae-Poetscher, van Zijl and Mori2004; Makris et al. Reference Makris, Kennedy, McInerney, Sorensen, Wang, Caviness and Pandya2005). Adolescents exposed to childhood maltreatment had significantly lower FA values in the bilateral superior longitudinal fasciculus, and those who developed a MDD at follow-up had the lowest values (Huang et al. Reference Huang, Gundapuneedi and Rao2012). WM integrity in the superior longitudinal fasciculus was consistently reported to be reduced in BD (Chaddock et al. Reference Chaddock, Barker, Marshall, Schulze, Hall, Fern, Walshe, Bramon, Chitnis, Murray and McDonald2009; Versace et al. Reference Versace, Almeida, Quevedo, Thompson, Terwilliger, Hassel, Kupfer and Phillips2010a ; Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ), with lithium increasing AD (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). Children with BD or with a first-degree relative with BD showed both a reduced FA in the superior longitudinal fasciculus compared with controls (Frazier et al. Reference Frazier, Breeze, Papadimitriou, Kennedy, Hodge, Moore, Howard, Rohan, Caviness and Makris2007), and in unaffected individuals at familial risk of mood disorder lower FA in the superior longitudinal fasciculus was associated with increasing genetic liability for BD (Chaddock et al. Reference Chaddock, Barker, Marshall, Schulze, Hall, Fern, Walshe, Bramon, Chitnis, Murray and McDonald2009) and with a higher polygenic risk allele load for MDD (Whalley et al. Reference Whalley, Sprooten, Hackett, Hall, Blackwood, Glahn, Bastin, Hall, Lawrie, Sussmann and McIntosh2013). Here we found our most significant peak effects of ACE on AD.
The inferior fronto-occipital fasciculus is a long association bundle that connects the inferior–lateral and dorsolateral prefrontal frontal cortex with posterior temporal and occipital cortices (Schmahmann et al. Reference Schmahmann, Pandya, Wang, Dai, D'Arceuil, de Crespigny and Wedeen2007), being thus involved in many brain functions. In healthy humans a reduced FA has been found to be associated with reduced inhibitory cognitive control during response selection (Forstmann et al. Reference Forstmann, Jahfari, Scholte, Wolfensteller, van den Wildenberg and Ridderinkhof2008). FA in the inferior fronto-occipital fasciculus is reduced during BD illness episodes (Zanetti et al. Reference Zanetti, Jackowski, Versace, Almeida, Hassel, Duran, Busatto, Kupfer and Phillips2009; Lu et al. Reference Lu, Zhou, Keedy, Reilly and Sweeney2011), and here lithium increases AD (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). Unaffected siblings had FA values in the inferior fronto-occipital fasciculus that were intermediate to and significantly different from those of healthy volunteers and patients with BD (Mahon et al. Reference Mahon, Burdick, Ikuta, Braga, Gruner, Malhotra and Szeszko2013), and, again, in unaffected relatives FA correlated with genetic risk (Chaddock et al. Reference Chaddock, Barker, Marshall, Schulze, Hall, Fern, Walshe, Bramon, Chitnis, Murray and McDonald2009).
Finally, the uncinate fasciculus connects the hippocampus, amygdala and anterior temporal lobe to the medial prefrontal cortex, thus allowing mood regulation and emotional processing, and playing a critical role in memory networks and in the comprehension and regulation of emotional responses to auditory stimuli (Schmahmann et al. Reference Schmahmann, Pandya, Wang, Dai, D'Arceuil, de Crespigny and Wedeen2007). Integrity of the uncinate fasciculus may facilitate language- and emotion-based evaluative processes as well as behavioural control functions that delay gratification (Olson et al. Reference Olson, Collins, Hooper, Muetzel, Lim and Luciana2009), is reduced in BD (McIntosh et al. Reference McIntosh, Munoz Maniega, Lymer, McKirdy, Hall, Sussmann, Bastin, Clayden, Johnstone and Lawrie2008; Sussmann et al. Reference Sussmann, Lymer, McKirdy, Moorhead, Munoz Maniega, Job, Hall, Bastin, Johnstone, Lawrie and McIntosh2009; Benedetti et al. Reference Benedetti, Absinta, Rocca, Radaelli, Poletti, Bernasconi, Dallaspezia, Pagani, Falini, Copetti, Colombo, Comi, Smeraldi and Filippi2011a ), and lithium treatment positively correlates with AD (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013). Moreover, in patients with BD, FA in the uncinate fasciculus positively correlates with the functional coupling between the amygdala and anterior cingulate cortex (Wang et al. Reference Wang, Kalmar, He, Jackowski, Chepenik, Edmiston, Tie, Gong, Shah, Jones, Uderman, Constable and Blumberg2009), which is a critical component of the circuits ensuring adequate emotional processing (Pezawas et al. Reference Pezawas, Meyer-Lindenberg, Drabant, Verchinski, Munoz, Kolachana, Egan, Mattay, Hariri and Weinberger2005).
These networks are crucial for cognitive functions known to be impaired in bipolar patients (Rubinsztein et al. Reference Rubinsztein, Michael, Paykel and Sahakian2000; Sole et al. Reference Sole, Bonnin, Torrent, Martinez-Aran, Popovic, Tabares-Seisdedos and Vieta2012). Early life stress has been found to reduce neurocognitive functioning (executive functioning, visual memory, spatial working memory) and emotional processing in healthy humans (Majer et al. Reference Majer, Nater, Lin, Capuron and Reeves2010; Gould et al. Reference Gould, Clarke, Heim, Harvey, Majer and Nemeroff2012). We then surmise that ACE could contribute to the impairment of cognitive and emotional processing in BD through the alteration of WM structure in specific networks. These effects of ACE are likely to be associated with other detrimental influences on WM integrity observed in BD. Increased genetic liability to familial BD has been associated with reduced FA across distributed regions of WM in patients and their unaffected relatives (Chaddock et al. Reference Chaddock, Barker, Marshall, Schulze, Hall, Fern, Walshe, Bramon, Chitnis, Murray and McDonald2009). Genetic factors affecting the severity of BD also influence WM integrity in BD (Benedetti et al. Reference Benedetti, Bollettini, Barberi, Radaelli, Poletti, Locatelli, Pirovano, Lorenzi, Falini, Colombo and Smeraldi2013), which is found to be reduced in the main tracts of the WM skeleton (Benedetti et al. Reference Benedetti, Yeh, Bellani, Radaelli, Nicoletti, Poletti, Falini, Dallaspezia, Colombo, Scotti, Smeraldi, Soares and Brambilla2011c ). Neuroinflammation has been associated with BD, with environmental stress, and with a disrupted integrity of myelin sheaths (Bartzokis, Reference Bartzokis2012). Gene × environment interactions, including ACE, could then play a major role in the observed relationship between BD and altered DTI measures of WM integrity.
However, the mechanisms by which this might happen remain elusive. The main limitation in interpreting in vivo DTI findings is that animal models of DTI measures have been mainly developed to test toxic effects of drugs or to model severe demyelinating diseases (Harsan et al. Reference Harsan, Poulet, Guignard, Steibel, Parizel, de Sousa, Boehm, Grucker and Ghandour2006), and not normal human neurodevelopment. The recently described trajectories of change in DTI measures over the human lifespan suggest non-linear relationships with age (Qiu et al. Reference Qiu, Tan, Zhou and Khong2008; Bartzokis et al. Reference Bartzokis, Lu, Heydari, Couvrette, Lee, Kalashyan, Freeman, Grinstead, Villablanca, Finn, Mintz, Alger and Altshuler2012; Lebel et al. Reference Lebel, Gee, Camicioli, Wieler, Martin and Beaulieu2012), and current models propose specific effects on brain structures depending on specific sensitive periods in which brain regions and WM tracts could be maximally susceptible to early stress (Andersen & Teicher, Reference Andersen and Teicher2008). However, prospective studies are still lacking and the available data do not include patients who developed neuropsychiatric diseases.
Given that proliferation and differentiation of oligodendrocytes, and the myelination of axons, are partly controlled by neurotransmitters (Karadottir & Attwell, Reference Karadottir and Attwell2007), factors affecting brain neurotransmitter function in response to stress could contribute to the association between early stress and reduced AD. One example can be found in a polymorphism in the promoter in the serotonin transporter (5-HTTLPR) (Canli & Lesch, Reference Canli and Lesch2007), which can independently influence both the behavioural effects of early stressors and WM structure in the general population: carriers of the low-activity alleles showed: (1) a higher association between stress and its worse consequences, including depression and suicide (Caspi et al. Reference Caspi, Hariri, Holmes, Uher and Moffitt2010); and (2) a reduced FA in the uncinate fasciculus (Pacheco et al. Reference Pacheco, Beevers, Benavides, McGeary, Stice and Schnyer2009). Other mechanisms possibly involved include inflammation, which is abnormally elevated in BD (Drexhage et al. Reference Drexhage, Knijff, Padmos, Heul-Nieuwenhuijzen, Beumer, Versnel and Drexhage2010), also prompted by environmental factors (Padmos et al. Reference Padmos, Van Baal, Vonk, Wijkhuijs, Kahn, Nolen and Drexhage2009), and can be increased by psychological distress (Raison et al. Reference Raison, Capuron and Miller2006). Many other genetic and epigenetic mechanisms can mediate the effects of stress (Dudley et al. Reference Dudley, Li, Kobor, Kippin and Bredy2011), but their effects on WM have not yet been investigated.
Strengths of the present study include a focused research question and state-of-the-art imaging methods. However, our results must be viewed in the light of several methodological limitations. The lack of a control group limits generalizability of the results, which, however, confirm previous literature in healthy subjects (see Introduction). The sample size allowed us to define the effects of ACE, but did not allow us to test their interaction with clinical variables, or their possible interactions with genetic or pharmacological factors. We measured no biological markers to test for possible mechanisms of the observed effects. Patients were non-drug-naive, and the drug treatments administered during the course of the illness could have influenced DTI measures, alone or interacting with the effects of early stress. Recruitment was in a single centre and in a single ethnic group, thus raising the possibility of population stratifications limiting the generalizability of the findings.
In conclusion, these limitations do not bias the main finding of an effect of ACE on WM microstructure in patients with BD, thus: (1) suggesting shared influences of stress on brain development of healthy and psychiatric populations, to be further confirmed and explored in enlarged and independent samples; and (2) confirming the usefulness of DTI for the study of the structural changes in the brain associated with BD.
Acknowledgements
The C.E.R.M.A.C. (Centro di Eccellenza Risonanza Magnetica ad Alto Campo) received research grants from the Italian Ministry of University and Scientific Research, from the Italian Ministry of Health, from the European Union (FP7 grant no. 222963 MOODINFLAME), from Trenta ore per la Vita Association, and from Janssen-Cilag.
Declaration of Interest
None.
