- Email: [email protected]
Association of resting-state network dysfunction with their dynamics of inter-network interactions in depression
Journal of Affective Disorders 174 (2015) 527–534
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research report
Association of resting-state network dysfunction with their dynamics of inter-network interactions in depression Maobin Wei a, Jiaolong Qin a, Rui Yan b, Kun Bi a, Chu Liu a, Zhijian Yao b,c,n, Qing Lu a,d,nn a
Key Laboratory of Child Development and Learning Science, Research Center of Learning Science, Southeast University, 2 sipailou, Nanjing 210096, China Academic Department of Psychiatry, Nanjing Brain Hospital, Nanjing Medical University, 249 Guangzhou Road, Nanjing 210029, China c Medical School, Nanjing University, 22 Hankou Road, Nanjing, 210093, China d Suzhou Research Institute of Southeast University, 399 Linquan Street, Suzhou, 215123, China b
art ic l e i nf o
a b s t r a c t
Article history: Received 21 April 2014 Received in revised form 25 October 2014 Accepted 4 December 2014 Available online 12 December 2014
Background: Network-level brain analysis on resting state has demonstrated that depression is not only associated with intra-network dysfunction, but relates to the disturbed interplay between the networks. However, the underlying associations between the intra-network dysfunction and the disturbed internetwork interactions remain unexplored. This study was aimed to explore the association of resting-state networks dysfunction with their dynamics of inter-network interactions in depression. Methods: Resting-state functional magnetic resonance imaging (fMRI) data were collected from 20 depressed patients and 20 matched healthy controls. We evaluated the Hurst exponents of the time series from resting-state networks, and employed multivariate pattern analysis to capture depressionassociated networks with increased or decreased Hurst values. Granger causalities between these networks were explored to undertake an intensive study of the dynamic inter-network interactions. Results: The default mode network (DMN) exhibited decreased Hurst value, indicative of more irregular oscillation within the DMN implicated in depressive symptoms. The ventromedial prefrontal network (vmPFN) and salience network (SN) with increased Hurst values, as compensatory mechanisms, continually enhanced the interactions to the DMN for trying hard to impel the DMN to function synchronously. On the other side, the DMN exerted frequently enhanced causality on the left frontoparietal network with elevated Hurst exponent, accompanied by imbalance between the fronto-parietal network and DMN circuits in depression. Limitations: This study suffers from small sample size and is confined to large-scale networks. Conclusions: Our preliminary findings mainly revealed the DMN-related dynamic interactions with the vmPFN, SN and the fronto-parietal network in depression, which might offer useful information for discovering the neuropathological mechanisms underlying the depressive symptoms. & 2014 Elsevier B.V. All rights reserved.
Keywords: Depression Resting-state networks Functional magnetic resonance imaging Hurst exponent Dynamic Granger causality
1. Introduction Pathophysiological observations in depression range from numerous regional anatomical and functional brain abnormalities to global, network-level brain dysfunction (Hamilton et al., 2013; Sheline et al., 2009; Tang et al., 2012; Wang et al., 2012). Largescale intrinsic brain network analysis of resting-state functional magnetic resonance imaging (fMRI) data has been increasingly employed in the studies of depression and other psychiatric disorders (Broyd et al., 2009; Orliac et al., 2013; Peng et al., 2012; White et al., 2010). Depression causes problems with many cognitive and emotion processes over the course of its occurrence. n
Corresponding author. Tel.: þ 86 25 82296252. Corresponding author. Tel.: þ 86 25 83795549. E-mail addresses: [email protected] (Z. Yao), [email protected] (Q. Lu).
nn
http://dx.doi.org/10.1016/j.jad.2014.12.020 0165-0327/& 2014 Elsevier B.V. All rights reserved.
It has been reported that ineffective transmission of information within the fronto-parietal network led to deficiency of cognition and executive functions (Brzezicka, 2013). The depressed patients exhibited increased functional connectivity within the taskpositive network including the fronto-parietal network implicated in attention and adaptive control, suggesting that the engraved negative emotional experiences might persist even in the absence of current external events (Zhou et al., 2010). Another important resting-state network (RSN), i.e. the default mode network (DMN), supports the self-reflective processes and its abnormality indicated a concurrence of negative attentional bias and negative memory bias in depression (Wang et al., 2012). The tendency of hyper-connectivity within the DMN was associated with psychological process such as rumination and brooding in depression (Berman et al., 2011; Zhu et al., 2012). In addition to the frontoparietal network and DMN, other RSNs, inclusive of the salience,
528
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
cerebellum, visual and affective networks, were considered to be disturbed to a certain extent with the depressive symptoms (Guo et al., 2013; Hamilton et al., 2013; Veer et al., 2010; Zeng et al., 2012). Recently, with the importance of the network in understanding the brain architecture and function, the increasing concern was drawn to the relations of inter-networks. It has been demonstrated that the efficient integration of information and emotion processing is involved in the coordinated work of a series of networks, even in resting-state (Di and Biswal, 2013a; Di et al., 2013; Doucet et al., 2011; Spreng et al., 2013; Yan and He, 2011). Further, the abnormal interactions or modulations of inter-networks were supposed to be associated with the mental disorders, especially schizophrenia. Schizophrenia patients exhibited increased correlation among the dominant RSNs with slightly more variability than healthy controls (Jafri et al., 2008). The salience network (SN) modulation of the fronto-parietal network and DMN was disrupted in schizophrenia and associated with cognitive deficits (Manoliu et al., 2013; Moran et al., 2013). In depression, anomalous functional connectivity between the amygdala and the dorsalateral prefrontal cortex might reflect the disrupted interactions between the SN and fronto-parietal network (Dannlowski et al., 2009; Lu et al., 2012). The fronto-insular cortex, when prompted by increased levels of DMN activity, initiated an adaptive engagement of the task-positive network, implying the modulation of the DMN to the task-positive network in depression (Hamilton et al., 2011b). It was also found that the depressed patients showed simultaneous dysfunction in the DMN, SN and executive network inferred from the elevated connectivity to these networks via the dorsal nexus (Sheline et al., 2010). All above evidence has demonstrated that intra-network dysfunction occurred along with the disturbed interplay between the RSNs in the depressive state, but a few reports associated the intra-network dysfunction with disturbed inter-network interactions. The underlying association remains unexplored. In this study, we hypothesized that anomalies of some networks were related to their disturbed interactions with the other networks in depression and the former tended to be the putative leading cause. To test this, we evaluated RSNs' Hurst exponents and dynamic causality interactions between networks, and further linked the dysfunction of intra- and internetworks to depression. Hurst exponent is a measure of selfsimilarity or regularity of a time series. That whether a RSN functions normally or not could be deduced from the Hurst exponent of its time series. Within the scope of 0.5–1, larger Hurst exponent means more regular fluctuation over time, suggesting a tendency of high coordinated signal organization within this network (Lai et al., 2010). Reports related to Hurst exponent analysis were documented in previous pathological and nonpathological studies, including Alzheimer's disease (Maxim et al., 2005), autism (Lai et al., 2010), normal aging (Wink et al., 2006)
and extraverted traits in human (Lei et al., 2013). In light of the definition of Hurst exponent implying a dynamic fluctuation over time of a certain time series, the Granger causality was computed over a sliding time-window to explore the network interactive dynamics in this study. The disturbed dynamic network interaction along with the abnormal intra-network fluctuation will be prospective to improve our understanding of the underlying pathophysiological mechanisms in the depressive disorder.
2. Materials and methods 2.1. Participants Participants comprised 20 healthy controls and 20 medicationnaïve major depressed individuals. The depressed patients aged from 20 to 45 were recruited from in-patient facilities at the hospital. Psychiatric diagnosis was based on DSM-IV criteria (APA, 2000) and Structure Clinical Interview for DSM-IV, and determined by an expert psychiatrist. All patients were rated on 17-item Hamilton Rating Scale for Depression (Hamilton, 1967) on the day of scanning and scored higher than 17. The matched healthy controls were recruited by print advertisements from the local community. All participants met the following inclusive criteria: (1) no current comorbidity with other major psychiatric, neurological or medical illness (e.g., learning disability, brain injury, psychotic symptoms, and bipolar disorder); (2) no history of dependency on or recent abuse of alcohol and/or drugs; (3) no physical contraindications for fMRI; (4) right-hand and native Chinese speakers. After a complete description of the study to all participants, written informed consent was obtained via forms approved by the Research Ethics Review Board. The demographic characteristics are summarized in Table 1. 2.2. Image acquisition All imaging data were acquired using a 3-T Siemens Verio scanner with an 8-channel radio frequency coil at the Brain Hospital affiliated to Nanjing Medical University in Nanjing, Jiangsu Province, China. Resting-state fMRI data were acquired using a gradient-echo EPI sequence with repetition time (TR)¼3000 ms, echo time (TE)¼40 ms, flip angle (FA)¼ 901, slice thickness¼4 mm, no slice gap, slices number¼32, field of view (FOV)¼ 240 240 mm2, matrix size¼ 64 64, in plane voxel resolution¼ 3.75 mm 3.75 mm, 6 min 45 s of scanning resulting in 133 volumes. Participants were instructed to keep their eyes closed, not to fall asleep and not to think of anything in particular during the scans. The T1-weighted structural images were acquired by magnetization-prepared rapid acquisition gradient-echo sequence
Table 1 Participant demographic and clinical characteristics.
Number of subjects Age, years (SD) Gender (male/female) Education, years (SD) HDRS, mean (SD) Cognitive disturbance factor in HDRS (SD) FD, mean (SD)
Depressed patients
Healthy controls
p-Value
20 34.3 (8.2) 10/10 13.9 (1.9) 25.8 (2.4) 4.0 (1.8) 0.102 (0.041)
20 30.8 (8.7) 6/14 15.0 (1.7) 1.4 (0.8)
0.20a 0.20b 0.059a
0.091 (0.028)
0.309a
SD, standard deviation; HDRS, Hamilton Depression Rating Scale; and FD, frame-wise displacement. a b
Two-tailed t-test. Pearson Chi-square two-tailed test.
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
with the following scan parameters: TR¼1900 ms, TE¼2.48 ms, FA ¼91, slices number¼ 176, slice thickness¼ 1 mm, voxel size¼ 1 1 1 mm3, FOV¼250 250 mm2 2.3. Preprocessing Resting-state fMRI data were preprocessed using SPM8 software package (http://www.fil.ion.ucl.ac.uk/spm/software/spm8). The first five functional volumes of each participant were discarded to account for T1 saturation effects and allow the participants' adaptation to the environment. The remaining 128 volumes were first corrected for different acquisition times of signals caused by interleaved fMRI acquisition. Then the motion correction was conducted using a least squares approach and a sixparameter (rigid body) linear transformation. No participant had head motion 4 3 mm of translation or 4 1.5 degree of rotation in any direction. An estimate of head motion at each time point was calculated as the frame-wise displacement (FD) using six displacements from rigid body motion correction procedure (Power et al., 2012). Two-sample t-test yielded no significant difference of FD between the depression patients and controls groups (Table 1). The structural image of each participant was co-registered to the mean functional image, and then transformed into standard Montreal Neurological Institute (MNI) space. Afterward, the functional images were spatially normalized to MNI space by applying the parameters of structural image normalization and resampled to 3 3 3 mm3 voxels. The resulting data were spatially smoothed using a 6 mm full width at half-maximum (FWHM) Gaussian kernel. 2.4. Group independent component analysis Using GIFT software (http://icatb.sourceforge.net), we applied group ICA to the preprocessed fMRI scans to decompose the resting-state data of the group into common spatiallyindependent components (Calhoun et al., 2009). The default number of 20 components as previously reported (Veer et al., 2010; Zuo et al., 2010) was used. The subject-specific spatial components and time courses were estimated by GICA3 back-reconstruction method. Statistical reliability of independent component decomposition was tested using the ICASSO toolbox (Himberg et al., 2004), and by running Infomax algorithm 20 times with different initial points. Twelve components were selected by visual inspection and template-match methods, potentially depicting functionally relevant RSNs (Kelly et al., 2010; Zuo et al., 2010). 2.5. Depression-associated RSNs 2.5.1. Hurst exponent calculation Rescaled Range (R/S) method was used to determine the Hurst exponent of each time series (Gilmore et al., 2002). The range (R) is the difference between the minimum and maximum cumulative of a portion of the time series at discrete integer-valued time over a time span m. For different m, the corresponding (R/S)m value is calculated via dividing the R by the standard deviation (S) of the values over the same portion of the time series. The slope of the regression of log(R/S)m on log(m) was taken as an estimator of the Hurst exponent. 2.5.2. Multivariate pattern analysis The Hurst exponents of the time courses corresponding to the 12 RSNs were analyzed by multivariate pattern analysis methods (Zeng et al., 2012) to capture depression-associated networks. The linear support vector machines (SVMs) with leave-out-one crossvalidation strategy were employed. The classification accuracy was
529
estimated and tested whether it was statistically significant or not via the permutation test (Wei et al., 2013). Given the two groups, depressed patients and healthy controls, with the lables þ1 and 1, the weight vectors were extracted and taken as the direction along which the training examples of both classes differ most (Mourão-Miranda et al., 2005). A positive value of weight vector item means relatively higher Hurst exponent in patients than in controls and a negative value means relatively higher Hurst exponent in controls than in patients (Ecker et al., 2010). The RSNs with the largest absolutes of weight vectors were considered as important networks (i.e. depression-associated networks) in distinguishing major depressive disorder (MDD) patients from healthy controls. The interactions between these networks were further explored via Granger causality analysis. 2.6. Conditional Granger causality analysis Sliding time-window conditional Granger causality methods (Luo et al., 2013) were used to assess dynamic interaction between these important networks. The window width was set to be 30 TRs and slid in steps of 1 TR, resulting in 99 windows. Within each window, conditional Granger causality analysis was conducted independently. We tookxas the time series of one RSN, y as that of another RSN, and z as all the remaining RSN time series other than x and y. The conditional Granger causality from time series y to x conditional on time series z was defined as F y-xjz . When a direct influence from time series y to x exists, the inclusion of past measurements of time series y in addition to that of time series x and z should result in better predictions of time series x, leading to F y-xjz 40. Otherwise, F y-xjz ¼ 0 means that no further improvement in the predication of time series x can be expected by including past measurements of time series y conditioned on the other time series z. The order of the autoregressive model was set to 1 as used in previous studies (Hamilton et al., 2011a; Liao et al., 2010). A permutation procedure was used to test the statistical significance of the computed conditional Granger causality in each window during between-group analysis. We randomized the labels of the depression patients and controls group by 10,000 times, and reassigned the two groups on the basis of the scrambled labels. The distribution of the differences of the mean Granger causality in two groups (patients subtract controls) after randomization was acquired for testing whether the Granger causality of the patients was significantly different from that of the healthy controls. The causality values under the threshold of distribution (po0.05) were not considered to be significant. The causalities satisfying the rightsided threshold demonstrated increased interactions in depression, while the causalities satisfying the left-sided threshold indicated increased interactions in healthy controls. After deriving the causality connectivity matrix in each window, the appearance frequencies of significant causalities for pair-wise RSNs over the time windows were estimated for further exploring on abnormal dynamic modulation of resting-state networks in depression. A summary of the approach is provided in Fig. 1.
3. Results 3.1. Depression-associated networks The SVM classifier achieved an accuracy of 90% (95% for patients, and 85% for healthy controls, p o 0.0001). According to the weight vectors obtained from the Hurst-based SVM method (Fig. 2), we found that the top weight elements were located in the right fronto-parietal network (RFPN), ventromedial prefrontal network (vmPFN), SN, DMN (posterior DMN and anterior DMN) and left fronto-parietal network (LFPN) in
530
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
Fig. 1. Illustration of analysis steps.
Fig. 2. Bar plot of the weight vectors of the resting-state networks in SVM classification. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
descending order. The RFPN and DMN exhibited decreased Hurst values in major depression (blue bar in Fig. 2), whereas the networks with increased Hurst values were anchored in the vmPFN, LFPN and SN (red bar in Fig. 2). The weight vectors of the two subnetworks of the DMN (i.e. posterior DMN and anterior DMN) were summed in that the vectors were negative values, demonstrating that the depressed patients exhibited decreased Hurst exponent within these two critical DMN subnetworks. These networks were then taken as the depression-associated networks with abnormal regular fluctuation. The vmPFN reported before mainly consisted of the anterior cingulate cortex, medial orbital frontal cortex, superior medial frontal cortex and caudate (Zuo et al., 2010). The LFPN and RFPN involving the inferior frontal cortex, medial frontal cortex, precuneus, inferior parietal cortex and angular gyrus were reported to be related to various high-order brain functions, including cognitive control, decision-making and emotional process. The DMN was divided into two sub-networks: the anterior default mode network (aDMN) mostly involving the medial frontal cortex, middle frontal cortex, and the anterior posterior cingulate cortex; the posterior default mode network (pDMN) covering the bilatearal precuneus, the posterior
cingulate cortex and bilateral angular gyrus. The SN was mainly involved in the anterior cingulate cortex, anterior insula and middle frontal cortex. The corresponding spatial component mappings are shown in Fig. 3. 3.2. Conditional Granger causality analysis To improve our understanding of the abnormal fluctuation of the signals of the depression-associated resting-state networks (Fig. 3) indicated by Hurst exponent in MDD, the sliding timewindow conditional Granger causality method was used to explore the dynamic interaction mechanism between these networks. Appearance frequencies of Granger causalities with group difference over all time windows are summarized in Fig. 4. The color of each cell denoted the appearance frequency of significant Granger causalities from the column network to the row network. The network pairs, whose causalities had relative higher frequencies, were considered to be involved greatly in inter-network interaction in resting-state and summarized in Fig. 5 for further analysis. Interactions were exclusively unidirectional among the six networks (Fig. 5). The SN and vmPFN exerted frequently significant influences on the pDMN with increased causal
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
531
Fig. 3. The spatial component mappings of the six networks. Each network is displayed according to radiological convention with the coordinates (x, y, z) shown in parentheses (MNI 152 standard space). vmPFN, ventromedial prefrontal network; LFPN, left fronto-parietal network; RFPN, right fronto-parietal network; aDMN, anterior default mode network; pDMN, posterior default mode network; and SN, salience network.
Fig. 4. The distribution matrix of the frequencies with significant Granger causalities appearing over all time windows. The left color bar represents the frequencies of appearance of causalities from the column network to the row network. vmPFN, ventromedial prefrontal network; LFPN, left fronto-parietal network; RFPN, right frontoparietal network; aDMN, anterior default mode network; pDMN, posterior default mode network; SN, salience network. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
regulation in MDD. The increased causality from the pDMN to the aDMN demonstrated abnormality of the DMN in depression. Within the fronto-parietal network, the RFPN directly caused enhanced influence on the LFPN. There also existed frequently abnormal interaction implied by increased causal modulation from the aDMN to the LPFN under depressive status.
3.3. Relationship to the clinical data To further explore the relationships of altered network connectivity to the clinical data, we took correlation analysis between the variances of the causalities of the pair-wise RSNs summarized in Fig. 5 and the scores of HAMD, and cognitive disturbance factor.
532
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
Fig. 5. The dynamical interactions between the six resting-state networks. The width of the arrows indicates the frequency of appearance of significant Granger causality. The yellow and blue ovals represent networks with higher and lower Hurst exponent in MDD respectively. vmPFN, ventromedial prefrontal network; LFPN, left fronto-parietal network; RFPN, right fronto-parietal network; aDMN, anterior default mode network; pDMN, posterior default mode network; SN, salience network. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
during rest than during task engagement, and also demonstrates persistent intra-network synchrony across a variety of states of consciousness and behavior (Sheline et al., 2010; Zhang and Raichle, 2010). In this study, the DMN was subdivided into the pDMN and aDMN. Both networks presented irregular intranetwork oscillation (negative weight vector of the Hurst-based SVM) in depression as compared with healthy controls, suggestive of abnormality of the DMN in depression. The irregularity might compel some regions within the default mode network to be hyperactive as a compensatory mechanism, which agreed with the previous reports demonstrating that the depressed patients are inclined to exhibit increased functional connectivity within the DMN (Zeng et al., 2012; Zhou et al., 2010), as well as the disturbed interactions between the subsystems within the DMN (Sambataro et al., 2013). Intriguingly, the variance of the causality from the pDMN to the aDMN was negatively correlated with the cognitive disturbance, implying dysfunctional internally focused cognitive style in depression (Marchetti et al., 2012). All the evidence leads to the conclusion that the depressive disorder is accompanied by dysfunction of the DMN. 4.2. DMN-related interactions
Fig. 6. Scatter plots of the negative relationship between the variance of the causality from the posterior default mode network (pDMN) to the anterior default mode network (aDMN) with the scores of the cognitive disturbance factor in depression.
The results showed no correlation between the variance of every pair-wise RSNs and the score of the HAMD. However, an intriguing result was thrown up, indicating that the variances, which described the dynamical change of the causality from the pDMN to the aDMN, were negatively correlated with the scores of the cognitive disturbance factor (r ¼ 0.64, p ¼0.0025, Fig. 6).
4. Discussion In the present study, we explored the dynamic interactions between resting-state networks with combinations of autocorrelation of these networks themselves. Hurst exponent, as a measure of autocorrelation of a time series, was estimated by R/S method, and taken as classification features to determine the networks assuming relatively importance in differentiating MDD patients from healthy controls. The depression-associated networks were located in the RFPN, vmPFN, SN, DMN and LFPN. These networks were made an intensive study of the dynamic interactions by sliding time-window Granger causality analysis. 4.1. Abnormal DMN The DMN has been posited to underlie function of innerdirected processes, such as introspection, self-referential thought, as well as supporting spontaneous cognition and monitoring the environment probably reflecting complementary and coexisting aspects of the conscious experience (Mantini and Vanduffel, 2013). The DMN is a task-negative network that exhibits more activity
The SN is involved in bottom-up detection of salience events, and plays a key role in switching between the DMN and executive network to mediate selection of physiologically relevant external and interoceptive signals (Menon and Uddin, 2010; Moran et al., 2013). The function of the vmPFN is closely related to emotion processing and cognitive control. It has been found that the depressed mood is correlated with increased activity within the vmPFN (Killgore and Yurgelun-Todd, 2006). When the DMN showed abnormality associated with the spontaneous negative bias and depressive rumination in MDD (Hamilton et al., 2011b; Marchetti et al., 2012), the depressed patients had difficulties in regulating self-focused thinking in order to engage in more goaldirected behavior at any time (Belleau et al., 2014), and the SN and vmPFN would exert dynamically enhanced interactions on the DMN at rest, trying to impel the DMN to function synchronously. These interesting interactions were also reflected by the results of Hurst exponent analysis that the DMN was possessed of reduced Hurst value, while the SN and vmPFN presented elevated Hurst values. The brain regions in DMN with a reduced Hurst value exhibited lower coordinated fluctuation, implying persistent negative thought in depression (Wei et al., 2013). In order to adjust the depressive status back to a normal level, the SN and vmPFN served as driving networks (Yan and He, 2011) and worked continuously to dispense from the negative bias by frequently enhanced intranetwork coordination. The fronto-parietal network was deemed to be a task-positive network, and had high priority in previous brain network's research, together with the DMN (Fox et al., 2005; Hamilton et al., 2011b). The fronto-parietal network plays a pivotal role in execution, control function, as well as emotion processing (Cole et al., 2013; Vincent et al., 2008). In depression, the anomaly of the DMN associated with persistent negative thought was implicated in the imbalance between the fronto-parietal network and DMN circuits. The frequently enhanced causality interaction from the DMN to the LFPN, which has been reported as the increased relapse risk in remitted depression (Jacobs et al., 2014), might imply the excessive and maladaptive self-focus and subsequent difficulties in switching to an extrospective perspective during rest (Marchetti et al., 2012). It has been reported that disturbances within the fronto-parietal network seem to be strongly associated with cognitive problems in depression, especially those concerning executive functions (Brzezicka, 2013). The cingulo-opercular system including the aDMN and SN and the fronto-parietal
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
network both seemed to maintain task-relevant information, however with the different mechanisms. When endogenous cognition, rumination or emotion happened, the cingulo-opercular system was obliged to provide stable regulation over the whole time, issuing the rapid adaptive control implemented by the fronto-parietal network (Dosenbach et al., 2008). Within the fronto-parietal network, it was also noted that the increased causality continually occurred from the RFPN to LFPN, probably resulting from the brain lateralization of emotional processing. Valence-lateralization theory postulates dominance of the left prefrontal cortex in positive emotions and the right prefrontal cortex in negative emotions (Hellige, 1993). The decreased Hurst exponent of the RFPN implied endogenous emotional instability and manifestation of negative bias in depressed patients. The LFPN with increased Hurst value was expected to activate in high intranetwork synchrony for inhibiting the negative bias, suggesting the frequent dynamic interaction of bilateral fronto-parietal network over the resting-state time. The SN, DMN and fronto-parietal network comprise a triple network and are considered as the core neurocognitive networks (Menon, 2011). Interplays within the triple network were demonstrated in most reports related to schizophrenia (Manoliu et al., 2014; Moran et al., 2013; Palaniyappan and Liddle, 2012; White et al., 2010). The SN not only modulated the activities of the DMN and fronto-parietal networks, but might also modulate the relationships between them (Di and Biswal, 2013b). In this study, it was found that the three core neurocognitive networks probably showed dysfunction implicated in frequently abnormal interactions within the triple network in depression (Manoliu et al., 2014). During the resting-state, the DMN accepted frequent enhanced interaction of the SN and exerted frequent enhanced interaction on the fronto-parietal network. These findings were instrumental in providing a deeper understanding of the depression-associated networks and therapeutic opportunities for depression. 4.3. Limitations Certain issues need to be considered in this study. First, relatively small sample size may limit the generalization of the preliminary results. Second, group ICA has been accepted as an advantageous method to extract resting-state brain network. However, the selection of model orders that determines the numbers of sub-network may be sensitive to different brain disorders. Some disorders may target specific brain system or regions. According to hierarchical modularity of human brain networks (Ferrarini et al., 2009), low model orders reflect the higher level of the hierarchy, while high model orders reflect the lower level of the hierarchy. Thus, an appropriate model order should be chosen cautiously for the goal of research. In this study, in light of the increased risk of false positives induced by high orders (Abou Elseoud et al., 2011), the number of ICs was determined to be 20 for detecting abnormalities of large-scale networks in MDD. The lower level of hierarchical brain network needs to be further explored for detecting specific sub-networks associated with depression.
5. Conclusion In summary, with the emerging dysfunction of DMN, the DMNrelated abnormal dynamic modulations with other RSNs happened, and the following was probably the disturbance of these RSNs. The SN and vmPFN exhibited high Hurst exponent for continually urging the DMN to dispend from the negative bias. The negative bias was accompanied by the dynamic interaction of the DMN on the LFPN. In
533
the resting-state, all RSNs constitute the whole brain functional system. Each network's dysfunction tended to be closely connected to abnormal dynamic interactions between networks. Our preliminary findings mainly revealed the DMN-related dynamic interactions with vmPFN, SN and the fronto-parietal network in depression, which might offer useful information for discovering the neuropathological mechanisms underlying the depressive symptoms.
Role of funding source This work was supported by the National Natural Science Foundation of China (81371522, 61372032), Jiangsu Clinical Medicine Technology Foundation (BL2012052), Jiangsu Natural Science Foundation (BK2012740) and the Fundamental Research Funds for the Central Universities (CXLX13_11).
Conflict of interest The authors declare that they have no conflict of interest.
Acknowledgments We thank Mr. Jianhuai Chen and Miss Lingling Hua who participated in the discussion of our results and gave some suggestions.
References Abou Elseoud, A., Littow, H., Remes, J., Starck, T., Nikkinen, J., Nissilä, J., Timonen, M., Tervonen, O., Kiviniemi, V., 2011. Group-ICA model order highlights patterns of functional brain connectivity. Front. Syst. Neurosci. 5, 00037. http://dx.doi.org/ 10.3389/fnsys.2011. APA, 2000. Diagnostic and Statistical Manual of Mental Disorders, 4th edition American Psychiatric Press, Washington, DC. Belleau, E.L., Taubitz, L.E., Larson, C.L., 2014. Imbalance of default mode and regulatory networks during externally focused processing in depression. Soc. Cogn. Affect. Neurosci., nsu117v1-nsu11. Berman, M.G., Peltier, S., Nee, D.E., Kross, E., Deldin, P.J., Jonides, J., 2011. Depression, rumination and the default network. Soc. Cogn. Affect. Neurosci. 6, 548–555. Broyd, S.J., Demanuele, C., Debener, S., Helps, S.K., James, C.J., Sonuga-Barke, E.J.S., 2009. Default-mode brain dysfunction in mental disorders: a systematic review. Neurosci. Biobehav. Rev. 33, 279–296. Brzezicka, A., 2013. Integrative deficits in depression and in negative mood states as a result of fronto-parietal network dysfunctions. Acta Neurobiol. Exp. (Wars) 73, 313–325. Calhoun, V.D., Liu, J., Adalı, T., 2009. A review of group ICA for fMRI data and ICA for joint inference of imaging, genetic, and ERP data. NeuroImage 45, S163–S172. Cole, M.W., Reynolds, J.R., Power, J.D., Repovs, G., Anticevic, A., Braver, T.S., 2013. Multi-task connectivity reveals flexible hubs for adaptive task control. Nat. Neurosci. 16, 1348–1355. Dannlowski, U., Ohrmann, P., Konrad, C., Domschke, K., Bauer, J., Kugel, H., Hohoff, C., Schöning, S., Kersting, A., Baune, B.T., Mortensen, L.S., Arolt, V., Zwitserlood, P., Deckert, J., Heindel, W., Suslow, T., 2009. Reduced amygdale-prefrontal coupling in major depression: association with MAOA genotype and illness severity. Int. J. Neuropsychopharmacol. 12, 11–22. Di, X., Biswal, B.B., 2013a. Modulatory interactions of resting-state brain functional connectivity. PLoS One 8, e71163. http://dx.doi.org/10.1371/journal.pone. 0071163. Di, X., Biswal, B.B., 2013b. Dynamic brain functional connectivity modulated by resting-state networks. Brain Struct. Funct., http://dx.doi.org/10.1007/s00429013-0634-3. Di, X., Gohel, S., Kim, E.H., Biswal, B.B., 2013. Task vs. rest-different network configurations between the coactivation and the Resting-state brain networks. Front. Hum. Neurosci. 7, http://dx.doi.org/10.3389/fnhum.2013.00493. Dosenbach, N.U.F., Fair, D.A., Cohen, A.L., Schlaggar, B.L., Petersen, S.E., 2008. A dualnetworks architecture of top-down control. Trends Cogn. Sci. 12, 99–105. Doucet, G., Naveau, M., Petit, L., Delcroix, N., Zago, L., Crivello, F., Jobard, G., TzourioMazoyer, N., Mazoyer, B., Mellet, E., Joliot, M., 2011. Brain activity at rest a multiscale hierarchical functional organization. J. Neurophysiol. 105, 2753–2763. Ecker, C., Rocha-Rego, V., Johnston, P., Mourao-Miranda, J., Marquand, A., Daly, E.M., Brammer, M.J., Murphy, C., Murphy, D.G., 2010. Investigating the predictive value of whole-brain structural MR scans in autism: a pattern classification approach. NeuroImage 49, 44–56. Ferrarini, L., Veer, I.M., Baerends, E., van Tol, M.J., Renken, R.J., van der Wee, N.J., Veltman, D.J., Aleman, A., Zitman, F.G., Penninx, B.W, van Buchem, M.A., Reiber, J.H., Rombouts, S.A., Milles, J., 2009. Hierarchical functional modularity in the resting-state human brain. Hum. Brain Mapp. 30, 2220–2231. Fox, M.D., Snyder, A.Z., Vincent, J.L., Corbetta, M., Van Essen, D.C., Raichle, M.E., 2005. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc. Natl. Acad. Sci. USA 102, 9673–9678.
534
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
Gilmore, M., Yu, C.X., Rhodes, T.L., 2002. Investigation of rescaled range analysis, the Hurst exponent, and long-time correlations in plasma turbulence. Phys. Plasmas 9, 1312–1317. Guo, W., Liu, F., Liu, J., Yu, L., Zhang, Z., Zhang, J., Chen, H., Xiao, C., 2013. Is there a cerebellar compensatory effort in first-episode, treatment-naive major depressive disorder at rest? Prog. Neuropsychopharmacol. Biol. Psychiatry 46, 13–18. Hamilton, J.P., Chen, G., Thomason, M.E., Schwartz, M.E., Gotlib, I.H., 2011a. Investigating neural primacy in major depressive disorder multivariate Granger causality analysis of resting-state fMRI time-series data. Mol. Psychiatry 16, 763–772. Hamilton, J.P., Chen, M.C., Gotlib, I.H., 2013. Neural systems approaches to understanding major depressive disorder: an intrinsic functional organization perspective. Neurobiol. Dis. 52, 4–11. Hamilton, J.P., Furman, D.J., Chang, C., Thomason, M.E., Dennis, E., Gotlib, I.H., 2011b. Default-mode and task-positive network activity in major depressive disorder: implications for adaptive and maladaptive rumination. Biol. Psychiatry 70, 327–333. Hamilton, M., 1967. Development of a rating scale for primary depressive illness. Br. J. Soc. Clin. Psychol. 6, 278–296. Hellige, J.B., 1993. Hemispheric Asymmetry: What's Right and What's Left. Harvard University Press, Cambridge, MA. Himberg, J., Hyvärinen, A., Esposito, F., 2004. Validating the independent components of neuroimaging time series via clustering and visualization. NeuroImage 22, 1214–1222. Jacobs, R.H., Jenkins, L.M., Gabriel, L.B., Barba, A., Ryan, K.A., Weisenbach, S.L., Verges, A., Baker, A.M., Peters, A.T., Crane, N.A., Gotlib, I.H., Zubieta, J.K., Phan, K. L., Langenecker, S.A., Welsh, R.C., 2014. Increased coupling of intrinsic networks in remitted depressed youth predicts rumination and cognitive control. PLoS One 9, e104366. Jafri, M.J., Pearlson, G.D., Stevens, M., Calhoun, V.D., 2008. A method for functional network connectivity among spatially independent resting-state components in schizophrenia. NeuroImage 39, 1666–1681. Kelly Jr., R.E., Alexopoulos, G.S., Wang, Z., Gunning, F.M., Murphy, C.F., Morimoto, S. S., Kanellopoulos, D., Jia, Z., Lim, K.O., Hoptman, M.J., 2010. Visual inspection of independent components: defining a procedure for artifact removal from fMRI data. J. Neurosci. Methods 189, 233–245. Killgore, W.D.S., Yurgelun-Todd, D.A., 2006. Ventromedial prefrontal activity correlates with depressed mood in adolescent children. Neuroreport 17, 167–171. Lai, M., Lombardo, M.V., Chakrabarti, B., Sadek, S.A., Pasco, G., Wheelwright, S.J., Bullmore, E.T., Baron-Cohen, S., Suckling, J., 2010. A shift to randomness of brain oscillations in people with autism. Biol. Psychiatry 68, 1092–1099. Lei, X., Zhao, Z., Chen, H., 2013. Extraversion is encoded by scale-free dynamics of default mode network. NeuroImage 74, 52–57. Liao, W., Mantini, D., Zhang, Z., Pan, Z., Ding, J., Gong, Q., Yang, Y., Chen, H., 2010. Evaluating the effective connectivity of resting state networks using conditional Granger causality. Biol. Cybern. 102, 57–69. Lu, Q., Li, H., Luo, G., Wang, Y., Tang, H., Han, L., Yao, Z., 2012. Impaired prefrontalamygdala effective connectivity is responsible for the dysfunction of emotion process in major depressive disorder: a dynamic causal modeling study on MEG. Neurosci. Lett. 523, 125–130. Luo, Q., Lu, W., Cheng, W., Valdes-Sosa, P.A., Wen, X., Ding, M., Feng, J., 2013. Spatiotemporal Granger causality: a new framework. NeuroImage 79, 241–263. Manoliu, A., Meng, C., Brandl, F., Doll, A., Tahmasian, M., Scherr, M., Schwerthöffer, D., Zimmer, C., Förstl, H., Bäuml, J., Riedl, V., Wohlschläger, A., Sorg, C., 2014. Insular dysfunction within the salience network is associated with severity of symptoms and aberrant inter-network connectivity in major depressive disorder. Front. Hum. Neurosci. 7, 930. Manoliu, A., Riedl, V., Zherdin, A., Mühlau, M., Schwerthöffer, D., Scherr, M., Peters, H., Zimmer, C., Förstl, H., Bäuml, J., Wohlschläger, A.M., Sorg, C., 2014. Aberrant dependence of default mode/central executive network interactions on anterior insular salience network activity in schizophrenia. Schizophr. Bull. 40, 428–437. Mantini, D., Vanduffel, W., 2013. Emerging roles of the brain's default network. Neuroscientist 19, 76–87. Marchetti, I., Koster, E.H., Sonuga-Barke, E.J., De Raedt, R., 2012. The default mode network and recurrent depression: a neurobiological model of cognitive risk factors. Neuropsychol. Rev. 22, 229–251. Maxim, V., Şendur, L., Fadili, J., Suckling, J., Gould, R., Howard, R., Bullmore, E., 2005. Fractional Gaussian noise, functional MRI and Alzheimer's disease. NeuroImage 25, 141–158. Menon, V., 2011. Large-scale brain networks and psychopathology: a unifying triple network model. Trends Cogn. Sci. 15, 483–506. Menon, V., Uddin, L., 2010. Saliency, switching, attention and control: a network model of insula function. Brain Struct. Funct. 214, 655–667. Moran, L.V., Tagamets, M.A., Sampath, H., Donnell, O., Stein, A., Kochunov, E.A., Hong, L.E., P., 2013. Disruption of anterior insula modulation of large-scale brain networks in schizophrenia. Biol. Psychiatry 74, 467–474.
Mourão-Miranda, J., Bokde, A.L.W., Born, C., Hampel, H., Stetter, M., 2005. Classifying brain states and determining the discriminating activation patterns: support vector machine on functional MRI data. NeuroImage 28, 980–995. Orliac, F., Naveau, M., Joliot, M., Delcroix, N., Razafimandimby, A., Brazo, P., Dollfus, S., Delamillieure, P., 2013. Links among resting-state default-mode network, salience network, and symptomatology in schizophrenia. Schizophr. Res. 148, 73–84. Palaniyappan, L., Liddle, P., 2012. Does the salience network play a cardinal role in psychosis? An emerging hypothesis of insular dysfunction. J. Psychiatry Neurosci. 37, 17–27. Peng, D., Shen, T., Zhang, J., Huang, J., Liu, J., Liu, S., Jiang, K., Xu, Y., Fang, Y., 2012. Abnormal functional connectivity with mood regulating circuit in unmedicated individual with major depression a resting-state functional magnetic resonance study. Chin. Med. J. (Engl.) 125, 3701–3706. Power, J.D., Barnes, K.A., Snyder, A.Z., Schlaggar, B.L., Petersen, S.E., 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. NeuroImage 59, 2142–2154. Sambataro, F., Wolf, N.D., Pennuto, M., Vasic, N., Wolf, R.C., 2013. Revisiting default mode network function in major depression: evidence for disrupted subsystem connectivity. Psychol. Med. 44, 2041–2051. Sheline, Y.I., Barch, D.M., Price, J.L., Rundle, M.M., Vaishnavi, S.N., Snyder, A.Z., Mintun, M.A., Wang, S., Coalson, R.S., Raichle, M.E., 2009. The default mode network and self-referential processes in depression. Proc. Natl. Acad. Sci. USA 106, 1942–1947. Sheline, Y.I., Price, J.L., Yan, Z., Mintun, M.A., 2010. Resting-state functional MRI in depression unmasks increased connectivity between networks via the dorsal nexus. Proc. Natl. Acad. Sci. USA 107, 11020–11025. Spreng, R.N., Sepulcre, J., Turner, G.R., Stevens, W.D., Schacter, D.L., 2013. Intrinsic architecture underlying the relations among the default, dorsal attention, and fronto-parietal control networks of the human brain. J. Cogn. Neurosci. 25, 74–86. Tang, Y., Kong, L., Wu, F., Womer, F., Jiang, W., Cao, Y., Ren, L., Wang, J., Fan, G., Blumberg, H.P., Xu, K., Wang, F., 2012. Decreased functional connectivity between the amygdala and the left ventral prefrontal cortex in treatmentnaive patients with major depressive disorder: a resting-state functional magnetic resonance imaging study. Psychol. Med. 43, 1921–1927. Veer, I.M., Beckmann, C., Van Tol, M., Ferrarini, L., Milles, J., Veltman, D., Aleman, A., Van Buchem, M.A., Van Der Wee, N.J.A., Rombouts, S.A.R., 2010. Whole brain resting-state analysis reveals decreased functional connectivity in major depression. Front. Syst. Neurosci. 4, http://dx.doi.org/10.3389/fnsys.2010.00041. Vincent, J.L., Kahn, I., Snyder, A.Z., Raichle, M.E., Buckner, R.L., 2008. Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. J. Neurophysiol. 100, 3328–3342. Wang, L., Hermens, D.F., Hickie, I.B., Lagopoulos, J., 2012. A systematic review of resting-state functional-MRI studies in major depression. J. Affect. Disord. 142, 6–12. Wei, M., Qin, J., Yan, R., Li, H., Yao, Z., Lu, Q., 2013. Identifying major depressive disorder using Hurst exponent of resting-state brain networks. Psychiatry Res. 214, 306–312. White, T.P., Joseph, V., Francis, S.T., Liddle, P.F., 2010. Aberrant salience network (bilateral insula and anterior cingulate cortex) connectivity during information processing in schizophrenia. Schizophr. Res. 123, 105–115. Wink, A.M., Bernard, F., Salvador, R., Bullmore, E., Suckling, J., 2006. Age and cholinergic effects on hemodynamics and functional coherence of human hippocampus. Neurobiol. Aging 27, 1395–1404. Yan, C., He, Y., 2011. Driving and driven architectures of directed small-world human brain functional networks. PLoS One 6, e23460. http://dx.doi.org/ 10.1371/journal.pone.0023460. Zeng, L., Shen, H., Liu, L., Wang, L., Li, B., Fang, P., Zhou, Z., Li, Y., Hu, D., 2012. Identifying major depression using whole-brain functional connectivity: a multivariate pattern analysis. Brain 135, 1498–1507. Zhang, D., Raichle, M.E., 2010. Disease and the brain's dark energy. Nat. Rev. Neurol. 6, 15–28. Zhou, Y., Yu, C., Zheng, H., Liu, Y., Song, M., Qin, W., Li, K., Jiang, T., 2010. Increased neural resources recruitment in the intrinsic organization in major depression. J. Affect. Disord. 121, 220–230. Zhu, X., Wang, X., Xiao, J., Liao, J., Zhong, M., Wang, W., Yao, S., 2012. Evidence of a dissociation pattern in resting-State default mode network connectivity in firstepisode, treatment-naive major depression patients. Biol. Psychiatry 71, 611–617. Zuo, X., Kelly, C., Adelstein, J.S., Klein, D.F., Castellanos, F.X., Milham, M.P., 2010. Reliable intrinsic connectivity networks: test–retest evaluation using ICA and dual regression approach. NeuroImage 49, 2163–2177.
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research report
Association of resting-state network dysfunction with their dynamics of inter-network interactions in depression Maobin Wei a, Jiaolong Qin a, Rui Yan b, Kun Bi a, Chu Liu a, Zhijian Yao b,c,n, Qing Lu a,d,nn a
Key Laboratory of Child Development and Learning Science, Research Center of Learning Science, Southeast University, 2 sipailou, Nanjing 210096, China Academic Department of Psychiatry, Nanjing Brain Hospital, Nanjing Medical University, 249 Guangzhou Road, Nanjing 210029, China c Medical School, Nanjing University, 22 Hankou Road, Nanjing, 210093, China d Suzhou Research Institute of Southeast University, 399 Linquan Street, Suzhou, 215123, China b
art ic l e i nf o
a b s t r a c t
Article history: Received 21 April 2014 Received in revised form 25 October 2014 Accepted 4 December 2014 Available online 12 December 2014
Background: Network-level brain analysis on resting state has demonstrated that depression is not only associated with intra-network dysfunction, but relates to the disturbed interplay between the networks. However, the underlying associations between the intra-network dysfunction and the disturbed internetwork interactions remain unexplored. This study was aimed to explore the association of resting-state networks dysfunction with their dynamics of inter-network interactions in depression. Methods: Resting-state functional magnetic resonance imaging (fMRI) data were collected from 20 depressed patients and 20 matched healthy controls. We evaluated the Hurst exponents of the time series from resting-state networks, and employed multivariate pattern analysis to capture depressionassociated networks with increased or decreased Hurst values. Granger causalities between these networks were explored to undertake an intensive study of the dynamic inter-network interactions. Results: The default mode network (DMN) exhibited decreased Hurst value, indicative of more irregular oscillation within the DMN implicated in depressive symptoms. The ventromedial prefrontal network (vmPFN) and salience network (SN) with increased Hurst values, as compensatory mechanisms, continually enhanced the interactions to the DMN for trying hard to impel the DMN to function synchronously. On the other side, the DMN exerted frequently enhanced causality on the left frontoparietal network with elevated Hurst exponent, accompanied by imbalance between the fronto-parietal network and DMN circuits in depression. Limitations: This study suffers from small sample size and is confined to large-scale networks. Conclusions: Our preliminary findings mainly revealed the DMN-related dynamic interactions with the vmPFN, SN and the fronto-parietal network in depression, which might offer useful information for discovering the neuropathological mechanisms underlying the depressive symptoms. & 2014 Elsevier B.V. All rights reserved.
Keywords: Depression Resting-state networks Functional magnetic resonance imaging Hurst exponent Dynamic Granger causality
1. Introduction Pathophysiological observations in depression range from numerous regional anatomical and functional brain abnormalities to global, network-level brain dysfunction (Hamilton et al., 2013; Sheline et al., 2009; Tang et al., 2012; Wang et al., 2012). Largescale intrinsic brain network analysis of resting-state functional magnetic resonance imaging (fMRI) data has been increasingly employed in the studies of depression and other psychiatric disorders (Broyd et al., 2009; Orliac et al., 2013; Peng et al., 2012; White et al., 2010). Depression causes problems with many cognitive and emotion processes over the course of its occurrence. n
Corresponding author. Tel.: þ 86 25 82296252. Corresponding author. Tel.: þ 86 25 83795549. E-mail addresses: [email protected] (Z. Yao), [email protected] (Q. Lu).
nn
http://dx.doi.org/10.1016/j.jad.2014.12.020 0165-0327/& 2014 Elsevier B.V. All rights reserved.
It has been reported that ineffective transmission of information within the fronto-parietal network led to deficiency of cognition and executive functions (Brzezicka, 2013). The depressed patients exhibited increased functional connectivity within the taskpositive network including the fronto-parietal network implicated in attention and adaptive control, suggesting that the engraved negative emotional experiences might persist even in the absence of current external events (Zhou et al., 2010). Another important resting-state network (RSN), i.e. the default mode network (DMN), supports the self-reflective processes and its abnormality indicated a concurrence of negative attentional bias and negative memory bias in depression (Wang et al., 2012). The tendency of hyper-connectivity within the DMN was associated with psychological process such as rumination and brooding in depression (Berman et al., 2011; Zhu et al., 2012). In addition to the frontoparietal network and DMN, other RSNs, inclusive of the salience,
528
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
cerebellum, visual and affective networks, were considered to be disturbed to a certain extent with the depressive symptoms (Guo et al., 2013; Hamilton et al., 2013; Veer et al., 2010; Zeng et al., 2012). Recently, with the importance of the network in understanding the brain architecture and function, the increasing concern was drawn to the relations of inter-networks. It has been demonstrated that the efficient integration of information and emotion processing is involved in the coordinated work of a series of networks, even in resting-state (Di and Biswal, 2013a; Di et al., 2013; Doucet et al., 2011; Spreng et al., 2013; Yan and He, 2011). Further, the abnormal interactions or modulations of inter-networks were supposed to be associated with the mental disorders, especially schizophrenia. Schizophrenia patients exhibited increased correlation among the dominant RSNs with slightly more variability than healthy controls (Jafri et al., 2008). The salience network (SN) modulation of the fronto-parietal network and DMN was disrupted in schizophrenia and associated with cognitive deficits (Manoliu et al., 2013; Moran et al., 2013). In depression, anomalous functional connectivity between the amygdala and the dorsalateral prefrontal cortex might reflect the disrupted interactions between the SN and fronto-parietal network (Dannlowski et al., 2009; Lu et al., 2012). The fronto-insular cortex, when prompted by increased levels of DMN activity, initiated an adaptive engagement of the task-positive network, implying the modulation of the DMN to the task-positive network in depression (Hamilton et al., 2011b). It was also found that the depressed patients showed simultaneous dysfunction in the DMN, SN and executive network inferred from the elevated connectivity to these networks via the dorsal nexus (Sheline et al., 2010). All above evidence has demonstrated that intra-network dysfunction occurred along with the disturbed interplay between the RSNs in the depressive state, but a few reports associated the intra-network dysfunction with disturbed inter-network interactions. The underlying association remains unexplored. In this study, we hypothesized that anomalies of some networks were related to their disturbed interactions with the other networks in depression and the former tended to be the putative leading cause. To test this, we evaluated RSNs' Hurst exponents and dynamic causality interactions between networks, and further linked the dysfunction of intra- and internetworks to depression. Hurst exponent is a measure of selfsimilarity or regularity of a time series. That whether a RSN functions normally or not could be deduced from the Hurst exponent of its time series. Within the scope of 0.5–1, larger Hurst exponent means more regular fluctuation over time, suggesting a tendency of high coordinated signal organization within this network (Lai et al., 2010). Reports related to Hurst exponent analysis were documented in previous pathological and nonpathological studies, including Alzheimer's disease (Maxim et al., 2005), autism (Lai et al., 2010), normal aging (Wink et al., 2006)
and extraverted traits in human (Lei et al., 2013). In light of the definition of Hurst exponent implying a dynamic fluctuation over time of a certain time series, the Granger causality was computed over a sliding time-window to explore the network interactive dynamics in this study. The disturbed dynamic network interaction along with the abnormal intra-network fluctuation will be prospective to improve our understanding of the underlying pathophysiological mechanisms in the depressive disorder.
2. Materials and methods 2.1. Participants Participants comprised 20 healthy controls and 20 medicationnaïve major depressed individuals. The depressed patients aged from 20 to 45 were recruited from in-patient facilities at the hospital. Psychiatric diagnosis was based on DSM-IV criteria (APA, 2000) and Structure Clinical Interview for DSM-IV, and determined by an expert psychiatrist. All patients were rated on 17-item Hamilton Rating Scale for Depression (Hamilton, 1967) on the day of scanning and scored higher than 17. The matched healthy controls were recruited by print advertisements from the local community. All participants met the following inclusive criteria: (1) no current comorbidity with other major psychiatric, neurological or medical illness (e.g., learning disability, brain injury, psychotic symptoms, and bipolar disorder); (2) no history of dependency on or recent abuse of alcohol and/or drugs; (3) no physical contraindications for fMRI; (4) right-hand and native Chinese speakers. After a complete description of the study to all participants, written informed consent was obtained via forms approved by the Research Ethics Review Board. The demographic characteristics are summarized in Table 1. 2.2. Image acquisition All imaging data were acquired using a 3-T Siemens Verio scanner with an 8-channel radio frequency coil at the Brain Hospital affiliated to Nanjing Medical University in Nanjing, Jiangsu Province, China. Resting-state fMRI data were acquired using a gradient-echo EPI sequence with repetition time (TR)¼3000 ms, echo time (TE)¼40 ms, flip angle (FA)¼ 901, slice thickness¼4 mm, no slice gap, slices number¼32, field of view (FOV)¼ 240 240 mm2, matrix size¼ 64 64, in plane voxel resolution¼ 3.75 mm 3.75 mm, 6 min 45 s of scanning resulting in 133 volumes. Participants were instructed to keep their eyes closed, not to fall asleep and not to think of anything in particular during the scans. The T1-weighted structural images were acquired by magnetization-prepared rapid acquisition gradient-echo sequence
Table 1 Participant demographic and clinical characteristics.
Number of subjects Age, years (SD) Gender (male/female) Education, years (SD) HDRS, mean (SD) Cognitive disturbance factor in HDRS (SD) FD, mean (SD)
Depressed patients
Healthy controls
p-Value
20 34.3 (8.2) 10/10 13.9 (1.9) 25.8 (2.4) 4.0 (1.8) 0.102 (0.041)
20 30.8 (8.7) 6/14 15.0 (1.7) 1.4 (0.8)
0.20a 0.20b 0.059a
0.091 (0.028)
0.309a
SD, standard deviation; HDRS, Hamilton Depression Rating Scale; and FD, frame-wise displacement. a b
Two-tailed t-test. Pearson Chi-square two-tailed test.
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
with the following scan parameters: TR¼1900 ms, TE¼2.48 ms, FA ¼91, slices number¼ 176, slice thickness¼ 1 mm, voxel size¼ 1 1 1 mm3, FOV¼250 250 mm2 2.3. Preprocessing Resting-state fMRI data were preprocessed using SPM8 software package (http://www.fil.ion.ucl.ac.uk/spm/software/spm8). The first five functional volumes of each participant were discarded to account for T1 saturation effects and allow the participants' adaptation to the environment. The remaining 128 volumes were first corrected for different acquisition times of signals caused by interleaved fMRI acquisition. Then the motion correction was conducted using a least squares approach and a sixparameter (rigid body) linear transformation. No participant had head motion 4 3 mm of translation or 4 1.5 degree of rotation in any direction. An estimate of head motion at each time point was calculated as the frame-wise displacement (FD) using six displacements from rigid body motion correction procedure (Power et al., 2012). Two-sample t-test yielded no significant difference of FD between the depression patients and controls groups (Table 1). The structural image of each participant was co-registered to the mean functional image, and then transformed into standard Montreal Neurological Institute (MNI) space. Afterward, the functional images were spatially normalized to MNI space by applying the parameters of structural image normalization and resampled to 3 3 3 mm3 voxels. The resulting data were spatially smoothed using a 6 mm full width at half-maximum (FWHM) Gaussian kernel. 2.4. Group independent component analysis Using GIFT software (http://icatb.sourceforge.net), we applied group ICA to the preprocessed fMRI scans to decompose the resting-state data of the group into common spatiallyindependent components (Calhoun et al., 2009). The default number of 20 components as previously reported (Veer et al., 2010; Zuo et al., 2010) was used. The subject-specific spatial components and time courses were estimated by GICA3 back-reconstruction method. Statistical reliability of independent component decomposition was tested using the ICASSO toolbox (Himberg et al., 2004), and by running Infomax algorithm 20 times with different initial points. Twelve components were selected by visual inspection and template-match methods, potentially depicting functionally relevant RSNs (Kelly et al., 2010; Zuo et al., 2010). 2.5. Depression-associated RSNs 2.5.1. Hurst exponent calculation Rescaled Range (R/S) method was used to determine the Hurst exponent of each time series (Gilmore et al., 2002). The range (R) is the difference between the minimum and maximum cumulative of a portion of the time series at discrete integer-valued time over a time span m. For different m, the corresponding (R/S)m value is calculated via dividing the R by the standard deviation (S) of the values over the same portion of the time series. The slope of the regression of log(R/S)m on log(m) was taken as an estimator of the Hurst exponent. 2.5.2. Multivariate pattern analysis The Hurst exponents of the time courses corresponding to the 12 RSNs were analyzed by multivariate pattern analysis methods (Zeng et al., 2012) to capture depression-associated networks. The linear support vector machines (SVMs) with leave-out-one crossvalidation strategy were employed. The classification accuracy was
529
estimated and tested whether it was statistically significant or not via the permutation test (Wei et al., 2013). Given the two groups, depressed patients and healthy controls, with the lables þ1 and 1, the weight vectors were extracted and taken as the direction along which the training examples of both classes differ most (Mourão-Miranda et al., 2005). A positive value of weight vector item means relatively higher Hurst exponent in patients than in controls and a negative value means relatively higher Hurst exponent in controls than in patients (Ecker et al., 2010). The RSNs with the largest absolutes of weight vectors were considered as important networks (i.e. depression-associated networks) in distinguishing major depressive disorder (MDD) patients from healthy controls. The interactions between these networks were further explored via Granger causality analysis. 2.6. Conditional Granger causality analysis Sliding time-window conditional Granger causality methods (Luo et al., 2013) were used to assess dynamic interaction between these important networks. The window width was set to be 30 TRs and slid in steps of 1 TR, resulting in 99 windows. Within each window, conditional Granger causality analysis was conducted independently. We tookxas the time series of one RSN, y as that of another RSN, and z as all the remaining RSN time series other than x and y. The conditional Granger causality from time series y to x conditional on time series z was defined as F y-xjz . When a direct influence from time series y to x exists, the inclusion of past measurements of time series y in addition to that of time series x and z should result in better predictions of time series x, leading to F y-xjz 40. Otherwise, F y-xjz ¼ 0 means that no further improvement in the predication of time series x can be expected by including past measurements of time series y conditioned on the other time series z. The order of the autoregressive model was set to 1 as used in previous studies (Hamilton et al., 2011a; Liao et al., 2010). A permutation procedure was used to test the statistical significance of the computed conditional Granger causality in each window during between-group analysis. We randomized the labels of the depression patients and controls group by 10,000 times, and reassigned the two groups on the basis of the scrambled labels. The distribution of the differences of the mean Granger causality in two groups (patients subtract controls) after randomization was acquired for testing whether the Granger causality of the patients was significantly different from that of the healthy controls. The causality values under the threshold of distribution (po0.05) were not considered to be significant. The causalities satisfying the rightsided threshold demonstrated increased interactions in depression, while the causalities satisfying the left-sided threshold indicated increased interactions in healthy controls. After deriving the causality connectivity matrix in each window, the appearance frequencies of significant causalities for pair-wise RSNs over the time windows were estimated for further exploring on abnormal dynamic modulation of resting-state networks in depression. A summary of the approach is provided in Fig. 1.
3. Results 3.1. Depression-associated networks The SVM classifier achieved an accuracy of 90% (95% for patients, and 85% for healthy controls, p o 0.0001). According to the weight vectors obtained from the Hurst-based SVM method (Fig. 2), we found that the top weight elements were located in the right fronto-parietal network (RFPN), ventromedial prefrontal network (vmPFN), SN, DMN (posterior DMN and anterior DMN) and left fronto-parietal network (LFPN) in
530
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
Fig. 1. Illustration of analysis steps.
Fig. 2. Bar plot of the weight vectors of the resting-state networks in SVM classification. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
descending order. The RFPN and DMN exhibited decreased Hurst values in major depression (blue bar in Fig. 2), whereas the networks with increased Hurst values were anchored in the vmPFN, LFPN and SN (red bar in Fig. 2). The weight vectors of the two subnetworks of the DMN (i.e. posterior DMN and anterior DMN) were summed in that the vectors were negative values, demonstrating that the depressed patients exhibited decreased Hurst exponent within these two critical DMN subnetworks. These networks were then taken as the depression-associated networks with abnormal regular fluctuation. The vmPFN reported before mainly consisted of the anterior cingulate cortex, medial orbital frontal cortex, superior medial frontal cortex and caudate (Zuo et al., 2010). The LFPN and RFPN involving the inferior frontal cortex, medial frontal cortex, precuneus, inferior parietal cortex and angular gyrus were reported to be related to various high-order brain functions, including cognitive control, decision-making and emotional process. The DMN was divided into two sub-networks: the anterior default mode network (aDMN) mostly involving the medial frontal cortex, middle frontal cortex, and the anterior posterior cingulate cortex; the posterior default mode network (pDMN) covering the bilatearal precuneus, the posterior
cingulate cortex and bilateral angular gyrus. The SN was mainly involved in the anterior cingulate cortex, anterior insula and middle frontal cortex. The corresponding spatial component mappings are shown in Fig. 3. 3.2. Conditional Granger causality analysis To improve our understanding of the abnormal fluctuation of the signals of the depression-associated resting-state networks (Fig. 3) indicated by Hurst exponent in MDD, the sliding timewindow conditional Granger causality method was used to explore the dynamic interaction mechanism between these networks. Appearance frequencies of Granger causalities with group difference over all time windows are summarized in Fig. 4. The color of each cell denoted the appearance frequency of significant Granger causalities from the column network to the row network. The network pairs, whose causalities had relative higher frequencies, were considered to be involved greatly in inter-network interaction in resting-state and summarized in Fig. 5 for further analysis. Interactions were exclusively unidirectional among the six networks (Fig. 5). The SN and vmPFN exerted frequently significant influences on the pDMN with increased causal
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
531
Fig. 3. The spatial component mappings of the six networks. Each network is displayed according to radiological convention with the coordinates (x, y, z) shown in parentheses (MNI 152 standard space). vmPFN, ventromedial prefrontal network; LFPN, left fronto-parietal network; RFPN, right fronto-parietal network; aDMN, anterior default mode network; pDMN, posterior default mode network; and SN, salience network.
Fig. 4. The distribution matrix of the frequencies with significant Granger causalities appearing over all time windows. The left color bar represents the frequencies of appearance of causalities from the column network to the row network. vmPFN, ventromedial prefrontal network; LFPN, left fronto-parietal network; RFPN, right frontoparietal network; aDMN, anterior default mode network; pDMN, posterior default mode network; SN, salience network. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
regulation in MDD. The increased causality from the pDMN to the aDMN demonstrated abnormality of the DMN in depression. Within the fronto-parietal network, the RFPN directly caused enhanced influence on the LFPN. There also existed frequently abnormal interaction implied by increased causal modulation from the aDMN to the LPFN under depressive status.
3.3. Relationship to the clinical data To further explore the relationships of altered network connectivity to the clinical data, we took correlation analysis between the variances of the causalities of the pair-wise RSNs summarized in Fig. 5 and the scores of HAMD, and cognitive disturbance factor.
532
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
Fig. 5. The dynamical interactions between the six resting-state networks. The width of the arrows indicates the frequency of appearance of significant Granger causality. The yellow and blue ovals represent networks with higher and lower Hurst exponent in MDD respectively. vmPFN, ventromedial prefrontal network; LFPN, left fronto-parietal network; RFPN, right fronto-parietal network; aDMN, anterior default mode network; pDMN, posterior default mode network; SN, salience network. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
during rest than during task engagement, and also demonstrates persistent intra-network synchrony across a variety of states of consciousness and behavior (Sheline et al., 2010; Zhang and Raichle, 2010). In this study, the DMN was subdivided into the pDMN and aDMN. Both networks presented irregular intranetwork oscillation (negative weight vector of the Hurst-based SVM) in depression as compared with healthy controls, suggestive of abnormality of the DMN in depression. The irregularity might compel some regions within the default mode network to be hyperactive as a compensatory mechanism, which agreed with the previous reports demonstrating that the depressed patients are inclined to exhibit increased functional connectivity within the DMN (Zeng et al., 2012; Zhou et al., 2010), as well as the disturbed interactions between the subsystems within the DMN (Sambataro et al., 2013). Intriguingly, the variance of the causality from the pDMN to the aDMN was negatively correlated with the cognitive disturbance, implying dysfunctional internally focused cognitive style in depression (Marchetti et al., 2012). All the evidence leads to the conclusion that the depressive disorder is accompanied by dysfunction of the DMN. 4.2. DMN-related interactions
Fig. 6. Scatter plots of the negative relationship between the variance of the causality from the posterior default mode network (pDMN) to the anterior default mode network (aDMN) with the scores of the cognitive disturbance factor in depression.
The results showed no correlation between the variance of every pair-wise RSNs and the score of the HAMD. However, an intriguing result was thrown up, indicating that the variances, which described the dynamical change of the causality from the pDMN to the aDMN, were negatively correlated with the scores of the cognitive disturbance factor (r ¼ 0.64, p ¼0.0025, Fig. 6).
4. Discussion In the present study, we explored the dynamic interactions between resting-state networks with combinations of autocorrelation of these networks themselves. Hurst exponent, as a measure of autocorrelation of a time series, was estimated by R/S method, and taken as classification features to determine the networks assuming relatively importance in differentiating MDD patients from healthy controls. The depression-associated networks were located in the RFPN, vmPFN, SN, DMN and LFPN. These networks were made an intensive study of the dynamic interactions by sliding time-window Granger causality analysis. 4.1. Abnormal DMN The DMN has been posited to underlie function of innerdirected processes, such as introspection, self-referential thought, as well as supporting spontaneous cognition and monitoring the environment probably reflecting complementary and coexisting aspects of the conscious experience (Mantini and Vanduffel, 2013). The DMN is a task-negative network that exhibits more activity
The SN is involved in bottom-up detection of salience events, and plays a key role in switching between the DMN and executive network to mediate selection of physiologically relevant external and interoceptive signals (Menon and Uddin, 2010; Moran et al., 2013). The function of the vmPFN is closely related to emotion processing and cognitive control. It has been found that the depressed mood is correlated with increased activity within the vmPFN (Killgore and Yurgelun-Todd, 2006). When the DMN showed abnormality associated with the spontaneous negative bias and depressive rumination in MDD (Hamilton et al., 2011b; Marchetti et al., 2012), the depressed patients had difficulties in regulating self-focused thinking in order to engage in more goaldirected behavior at any time (Belleau et al., 2014), and the SN and vmPFN would exert dynamically enhanced interactions on the DMN at rest, trying to impel the DMN to function synchronously. These interesting interactions were also reflected by the results of Hurst exponent analysis that the DMN was possessed of reduced Hurst value, while the SN and vmPFN presented elevated Hurst values. The brain regions in DMN with a reduced Hurst value exhibited lower coordinated fluctuation, implying persistent negative thought in depression (Wei et al., 2013). In order to adjust the depressive status back to a normal level, the SN and vmPFN served as driving networks (Yan and He, 2011) and worked continuously to dispense from the negative bias by frequently enhanced intranetwork coordination. The fronto-parietal network was deemed to be a task-positive network, and had high priority in previous brain network's research, together with the DMN (Fox et al., 2005; Hamilton et al., 2011b). The fronto-parietal network plays a pivotal role in execution, control function, as well as emotion processing (Cole et al., 2013; Vincent et al., 2008). In depression, the anomaly of the DMN associated with persistent negative thought was implicated in the imbalance between the fronto-parietal network and DMN circuits. The frequently enhanced causality interaction from the DMN to the LFPN, which has been reported as the increased relapse risk in remitted depression (Jacobs et al., 2014), might imply the excessive and maladaptive self-focus and subsequent difficulties in switching to an extrospective perspective during rest (Marchetti et al., 2012). It has been reported that disturbances within the fronto-parietal network seem to be strongly associated with cognitive problems in depression, especially those concerning executive functions (Brzezicka, 2013). The cingulo-opercular system including the aDMN and SN and the fronto-parietal
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
network both seemed to maintain task-relevant information, however with the different mechanisms. When endogenous cognition, rumination or emotion happened, the cingulo-opercular system was obliged to provide stable regulation over the whole time, issuing the rapid adaptive control implemented by the fronto-parietal network (Dosenbach et al., 2008). Within the fronto-parietal network, it was also noted that the increased causality continually occurred from the RFPN to LFPN, probably resulting from the brain lateralization of emotional processing. Valence-lateralization theory postulates dominance of the left prefrontal cortex in positive emotions and the right prefrontal cortex in negative emotions (Hellige, 1993). The decreased Hurst exponent of the RFPN implied endogenous emotional instability and manifestation of negative bias in depressed patients. The LFPN with increased Hurst value was expected to activate in high intranetwork synchrony for inhibiting the negative bias, suggesting the frequent dynamic interaction of bilateral fronto-parietal network over the resting-state time. The SN, DMN and fronto-parietal network comprise a triple network and are considered as the core neurocognitive networks (Menon, 2011). Interplays within the triple network were demonstrated in most reports related to schizophrenia (Manoliu et al., 2014; Moran et al., 2013; Palaniyappan and Liddle, 2012; White et al., 2010). The SN not only modulated the activities of the DMN and fronto-parietal networks, but might also modulate the relationships between them (Di and Biswal, 2013b). In this study, it was found that the three core neurocognitive networks probably showed dysfunction implicated in frequently abnormal interactions within the triple network in depression (Manoliu et al., 2014). During the resting-state, the DMN accepted frequent enhanced interaction of the SN and exerted frequent enhanced interaction on the fronto-parietal network. These findings were instrumental in providing a deeper understanding of the depression-associated networks and therapeutic opportunities for depression. 4.3. Limitations Certain issues need to be considered in this study. First, relatively small sample size may limit the generalization of the preliminary results. Second, group ICA has been accepted as an advantageous method to extract resting-state brain network. However, the selection of model orders that determines the numbers of sub-network may be sensitive to different brain disorders. Some disorders may target specific brain system or regions. According to hierarchical modularity of human brain networks (Ferrarini et al., 2009), low model orders reflect the higher level of the hierarchy, while high model orders reflect the lower level of the hierarchy. Thus, an appropriate model order should be chosen cautiously for the goal of research. In this study, in light of the increased risk of false positives induced by high orders (Abou Elseoud et al., 2011), the number of ICs was determined to be 20 for detecting abnormalities of large-scale networks in MDD. The lower level of hierarchical brain network needs to be further explored for detecting specific sub-networks associated with depression.
5. Conclusion In summary, with the emerging dysfunction of DMN, the DMNrelated abnormal dynamic modulations with other RSNs happened, and the following was probably the disturbance of these RSNs. The SN and vmPFN exhibited high Hurst exponent for continually urging the DMN to dispend from the negative bias. The negative bias was accompanied by the dynamic interaction of the DMN on the LFPN. In
533
the resting-state, all RSNs constitute the whole brain functional system. Each network's dysfunction tended to be closely connected to abnormal dynamic interactions between networks. Our preliminary findings mainly revealed the DMN-related dynamic interactions with vmPFN, SN and the fronto-parietal network in depression, which might offer useful information for discovering the neuropathological mechanisms underlying the depressive symptoms.
Role of funding source This work was supported by the National Natural Science Foundation of China (81371522, 61372032), Jiangsu Clinical Medicine Technology Foundation (BL2012052), Jiangsu Natural Science Foundation (BK2012740) and the Fundamental Research Funds for the Central Universities (CXLX13_11).
Conflict of interest The authors declare that they have no conflict of interest.
Acknowledgments We thank Mr. Jianhuai Chen and Miss Lingling Hua who participated in the discussion of our results and gave some suggestions.
References Abou Elseoud, A., Littow, H., Remes, J., Starck, T., Nikkinen, J., Nissilä, J., Timonen, M., Tervonen, O., Kiviniemi, V., 2011. Group-ICA model order highlights patterns of functional brain connectivity. Front. Syst. Neurosci. 5, 00037. http://dx.doi.org/ 10.3389/fnsys.2011. APA, 2000. Diagnostic and Statistical Manual of Mental Disorders, 4th edition American Psychiatric Press, Washington, DC. Belleau, E.L., Taubitz, L.E., Larson, C.L., 2014. Imbalance of default mode and regulatory networks during externally focused processing in depression. Soc. Cogn. Affect. Neurosci., nsu117v1-nsu11. Berman, M.G., Peltier, S., Nee, D.E., Kross, E., Deldin, P.J., Jonides, J., 2011. Depression, rumination and the default network. Soc. Cogn. Affect. Neurosci. 6, 548–555. Broyd, S.J., Demanuele, C., Debener, S., Helps, S.K., James, C.J., Sonuga-Barke, E.J.S., 2009. Default-mode brain dysfunction in mental disorders: a systematic review. Neurosci. Biobehav. Rev. 33, 279–296. Brzezicka, A., 2013. Integrative deficits in depression and in negative mood states as a result of fronto-parietal network dysfunctions. Acta Neurobiol. Exp. (Wars) 73, 313–325. Calhoun, V.D., Liu, J., Adalı, T., 2009. A review of group ICA for fMRI data and ICA for joint inference of imaging, genetic, and ERP data. NeuroImage 45, S163–S172. Cole, M.W., Reynolds, J.R., Power, J.D., Repovs, G., Anticevic, A., Braver, T.S., 2013. Multi-task connectivity reveals flexible hubs for adaptive task control. Nat. Neurosci. 16, 1348–1355. Dannlowski, U., Ohrmann, P., Konrad, C., Domschke, K., Bauer, J., Kugel, H., Hohoff, C., Schöning, S., Kersting, A., Baune, B.T., Mortensen, L.S., Arolt, V., Zwitserlood, P., Deckert, J., Heindel, W., Suslow, T., 2009. Reduced amygdale-prefrontal coupling in major depression: association with MAOA genotype and illness severity. Int. J. Neuropsychopharmacol. 12, 11–22. Di, X., Biswal, B.B., 2013a. Modulatory interactions of resting-state brain functional connectivity. PLoS One 8, e71163. http://dx.doi.org/10.1371/journal.pone. 0071163. Di, X., Biswal, B.B., 2013b. Dynamic brain functional connectivity modulated by resting-state networks. Brain Struct. Funct., http://dx.doi.org/10.1007/s00429013-0634-3. Di, X., Gohel, S., Kim, E.H., Biswal, B.B., 2013. Task vs. rest-different network configurations between the coactivation and the Resting-state brain networks. Front. Hum. Neurosci. 7, http://dx.doi.org/10.3389/fnhum.2013.00493. Dosenbach, N.U.F., Fair, D.A., Cohen, A.L., Schlaggar, B.L., Petersen, S.E., 2008. A dualnetworks architecture of top-down control. Trends Cogn. Sci. 12, 99–105. Doucet, G., Naveau, M., Petit, L., Delcroix, N., Zago, L., Crivello, F., Jobard, G., TzourioMazoyer, N., Mazoyer, B., Mellet, E., Joliot, M., 2011. Brain activity at rest a multiscale hierarchical functional organization. J. Neurophysiol. 105, 2753–2763. Ecker, C., Rocha-Rego, V., Johnston, P., Mourao-Miranda, J., Marquand, A., Daly, E.M., Brammer, M.J., Murphy, C., Murphy, D.G., 2010. Investigating the predictive value of whole-brain structural MR scans in autism: a pattern classification approach. NeuroImage 49, 44–56. Ferrarini, L., Veer, I.M., Baerends, E., van Tol, M.J., Renken, R.J., van der Wee, N.J., Veltman, D.J., Aleman, A., Zitman, F.G., Penninx, B.W, van Buchem, M.A., Reiber, J.H., Rombouts, S.A., Milles, J., 2009. Hierarchical functional modularity in the resting-state human brain. Hum. Brain Mapp. 30, 2220–2231. Fox, M.D., Snyder, A.Z., Vincent, J.L., Corbetta, M., Van Essen, D.C., Raichle, M.E., 2005. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc. Natl. Acad. Sci. USA 102, 9673–9678.
534
M. Wei et al. / Journal of Affective Disorders 174 (2015) 527–534
Gilmore, M., Yu, C.X., Rhodes, T.L., 2002. Investigation of rescaled range analysis, the Hurst exponent, and long-time correlations in plasma turbulence. Phys. Plasmas 9, 1312–1317. Guo, W., Liu, F., Liu, J., Yu, L., Zhang, Z., Zhang, J., Chen, H., Xiao, C., 2013. Is there a cerebellar compensatory effort in first-episode, treatment-naive major depressive disorder at rest? Prog. Neuropsychopharmacol. Biol. Psychiatry 46, 13–18. Hamilton, J.P., Chen, G., Thomason, M.E., Schwartz, M.E., Gotlib, I.H., 2011a. Investigating neural primacy in major depressive disorder multivariate Granger causality analysis of resting-state fMRI time-series data. Mol. Psychiatry 16, 763–772. Hamilton, J.P., Chen, M.C., Gotlib, I.H., 2013. Neural systems approaches to understanding major depressive disorder: an intrinsic functional organization perspective. Neurobiol. Dis. 52, 4–11. Hamilton, J.P., Furman, D.J., Chang, C., Thomason, M.E., Dennis, E., Gotlib, I.H., 2011b. Default-mode and task-positive network activity in major depressive disorder: implications for adaptive and maladaptive rumination. Biol. Psychiatry 70, 327–333. Hamilton, M., 1967. Development of a rating scale for primary depressive illness. Br. J. Soc. Clin. Psychol. 6, 278–296. Hellige, J.B., 1993. Hemispheric Asymmetry: What's Right and What's Left. Harvard University Press, Cambridge, MA. Himberg, J., Hyvärinen, A., Esposito, F., 2004. Validating the independent components of neuroimaging time series via clustering and visualization. NeuroImage 22, 1214–1222. Jacobs, R.H., Jenkins, L.M., Gabriel, L.B., Barba, A., Ryan, K.A., Weisenbach, S.L., Verges, A., Baker, A.M., Peters, A.T., Crane, N.A., Gotlib, I.H., Zubieta, J.K., Phan, K. L., Langenecker, S.A., Welsh, R.C., 2014. Increased coupling of intrinsic networks in remitted depressed youth predicts rumination and cognitive control. PLoS One 9, e104366. Jafri, M.J., Pearlson, G.D., Stevens, M., Calhoun, V.D., 2008. A method for functional network connectivity among spatially independent resting-state components in schizophrenia. NeuroImage 39, 1666–1681. Kelly Jr., R.E., Alexopoulos, G.S., Wang, Z., Gunning, F.M., Murphy, C.F., Morimoto, S. S., Kanellopoulos, D., Jia, Z., Lim, K.O., Hoptman, M.J., 2010. Visual inspection of independent components: defining a procedure for artifact removal from fMRI data. J. Neurosci. Methods 189, 233–245. Killgore, W.D.S., Yurgelun-Todd, D.A., 2006. Ventromedial prefrontal activity correlates with depressed mood in adolescent children. Neuroreport 17, 167–171. Lai, M., Lombardo, M.V., Chakrabarti, B., Sadek, S.A., Pasco, G., Wheelwright, S.J., Bullmore, E.T., Baron-Cohen, S., Suckling, J., 2010. A shift to randomness of brain oscillations in people with autism. Biol. Psychiatry 68, 1092–1099. Lei, X., Zhao, Z., Chen, H., 2013. Extraversion is encoded by scale-free dynamics of default mode network. NeuroImage 74, 52–57. Liao, W., Mantini, D., Zhang, Z., Pan, Z., Ding, J., Gong, Q., Yang, Y., Chen, H., 2010. Evaluating the effective connectivity of resting state networks using conditional Granger causality. Biol. Cybern. 102, 57–69. Lu, Q., Li, H., Luo, G., Wang, Y., Tang, H., Han, L., Yao, Z., 2012. Impaired prefrontalamygdala effective connectivity is responsible for the dysfunction of emotion process in major depressive disorder: a dynamic causal modeling study on MEG. Neurosci. Lett. 523, 125–130. Luo, Q., Lu, W., Cheng, W., Valdes-Sosa, P.A., Wen, X., Ding, M., Feng, J., 2013. Spatiotemporal Granger causality: a new framework. NeuroImage 79, 241–263. Manoliu, A., Meng, C., Brandl, F., Doll, A., Tahmasian, M., Scherr, M., Schwerthöffer, D., Zimmer, C., Förstl, H., Bäuml, J., Riedl, V., Wohlschläger, A., Sorg, C., 2014. Insular dysfunction within the salience network is associated with severity of symptoms and aberrant inter-network connectivity in major depressive disorder. Front. Hum. Neurosci. 7, 930. Manoliu, A., Riedl, V., Zherdin, A., Mühlau, M., Schwerthöffer, D., Scherr, M., Peters, H., Zimmer, C., Förstl, H., Bäuml, J., Wohlschläger, A.M., Sorg, C., 2014. Aberrant dependence of default mode/central executive network interactions on anterior insular salience network activity in schizophrenia. Schizophr. Bull. 40, 428–437. Mantini, D., Vanduffel, W., 2013. Emerging roles of the brain's default network. Neuroscientist 19, 76–87. Marchetti, I., Koster, E.H., Sonuga-Barke, E.J., De Raedt, R., 2012. The default mode network and recurrent depression: a neurobiological model of cognitive risk factors. Neuropsychol. Rev. 22, 229–251. Maxim, V., Şendur, L., Fadili, J., Suckling, J., Gould, R., Howard, R., Bullmore, E., 2005. Fractional Gaussian noise, functional MRI and Alzheimer's disease. NeuroImage 25, 141–158. Menon, V., 2011. Large-scale brain networks and psychopathology: a unifying triple network model. Trends Cogn. Sci. 15, 483–506. Menon, V., Uddin, L., 2010. Saliency, switching, attention and control: a network model of insula function. Brain Struct. Funct. 214, 655–667. Moran, L.V., Tagamets, M.A., Sampath, H., Donnell, O., Stein, A., Kochunov, E.A., Hong, L.E., P., 2013. Disruption of anterior insula modulation of large-scale brain networks in schizophrenia. Biol. Psychiatry 74, 467–474.
Mourão-Miranda, J., Bokde, A.L.W., Born, C., Hampel, H., Stetter, M., 2005. Classifying brain states and determining the discriminating activation patterns: support vector machine on functional MRI data. NeuroImage 28, 980–995. Orliac, F., Naveau, M., Joliot, M., Delcroix, N., Razafimandimby, A., Brazo, P., Dollfus, S., Delamillieure, P., 2013. Links among resting-state default-mode network, salience network, and symptomatology in schizophrenia. Schizophr. Res. 148, 73–84. Palaniyappan, L., Liddle, P., 2012. Does the salience network play a cardinal role in psychosis? An emerging hypothesis of insular dysfunction. J. Psychiatry Neurosci. 37, 17–27. Peng, D., Shen, T., Zhang, J., Huang, J., Liu, J., Liu, S., Jiang, K., Xu, Y., Fang, Y., 2012. Abnormal functional connectivity with mood regulating circuit in unmedicated individual with major depression a resting-state functional magnetic resonance study. Chin. Med. J. (Engl.) 125, 3701–3706. Power, J.D., Barnes, K.A., Snyder, A.Z., Schlaggar, B.L., Petersen, S.E., 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. NeuroImage 59, 2142–2154. Sambataro, F., Wolf, N.D., Pennuto, M., Vasic, N., Wolf, R.C., 2013. Revisiting default mode network function in major depression: evidence for disrupted subsystem connectivity. Psychol. Med. 44, 2041–2051. Sheline, Y.I., Barch, D.M., Price, J.L., Rundle, M.M., Vaishnavi, S.N., Snyder, A.Z., Mintun, M.A., Wang, S., Coalson, R.S., Raichle, M.E., 2009. The default mode network and self-referential processes in depression. Proc. Natl. Acad. Sci. USA 106, 1942–1947. Sheline, Y.I., Price, J.L., Yan, Z., Mintun, M.A., 2010. Resting-state functional MRI in depression unmasks increased connectivity between networks via the dorsal nexus. Proc. Natl. Acad. Sci. USA 107, 11020–11025. Spreng, R.N., Sepulcre, J., Turner, G.R., Stevens, W.D., Schacter, D.L., 2013. Intrinsic architecture underlying the relations among the default, dorsal attention, and fronto-parietal control networks of the human brain. J. Cogn. Neurosci. 25, 74–86. Tang, Y., Kong, L., Wu, F., Womer, F., Jiang, W., Cao, Y., Ren, L., Wang, J., Fan, G., Blumberg, H.P., Xu, K., Wang, F., 2012. Decreased functional connectivity between the amygdala and the left ventral prefrontal cortex in treatmentnaive patients with major depressive disorder: a resting-state functional magnetic resonance imaging study. Psychol. Med. 43, 1921–1927. Veer, I.M., Beckmann, C., Van Tol, M., Ferrarini, L., Milles, J., Veltman, D., Aleman, A., Van Buchem, M.A., Van Der Wee, N.J.A., Rombouts, S.A.R., 2010. Whole brain resting-state analysis reveals decreased functional connectivity in major depression. Front. Syst. Neurosci. 4, http://dx.doi.org/10.3389/fnsys.2010.00041. Vincent, J.L., Kahn, I., Snyder, A.Z., Raichle, M.E., Buckner, R.L., 2008. Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. J. Neurophysiol. 100, 3328–3342. Wang, L., Hermens, D.F., Hickie, I.B., Lagopoulos, J., 2012. A systematic review of resting-state functional-MRI studies in major depression. J. Affect. Disord. 142, 6–12. Wei, M., Qin, J., Yan, R., Li, H., Yao, Z., Lu, Q., 2013. Identifying major depressive disorder using Hurst exponent of resting-state brain networks. Psychiatry Res. 214, 306–312. White, T.P., Joseph, V., Francis, S.T., Liddle, P.F., 2010. Aberrant salience network (bilateral insula and anterior cingulate cortex) connectivity during information processing in schizophrenia. Schizophr. Res. 123, 105–115. Wink, A.M., Bernard, F., Salvador, R., Bullmore, E., Suckling, J., 2006. Age and cholinergic effects on hemodynamics and functional coherence of human hippocampus. Neurobiol. Aging 27, 1395–1404. Yan, C., He, Y., 2011. Driving and driven architectures of directed small-world human brain functional networks. PLoS One 6, e23460. http://dx.doi.org/ 10.1371/journal.pone.0023460. Zeng, L., Shen, H., Liu, L., Wang, L., Li, B., Fang, P., Zhou, Z., Li, Y., Hu, D., 2012. Identifying major depression using whole-brain functional connectivity: a multivariate pattern analysis. Brain 135, 1498–1507. Zhang, D., Raichle, M.E., 2010. Disease and the brain's dark energy. Nat. Rev. Neurol. 6, 15–28. Zhou, Y., Yu, C., Zheng, H., Liu, Y., Song, M., Qin, W., Li, K., Jiang, T., 2010. Increased neural resources recruitment in the intrinsic organization in major depression. J. Affect. Disord. 121, 220–230. Zhu, X., Wang, X., Xiao, J., Liao, J., Zhong, M., Wang, W., Yao, S., 2012. Evidence of a dissociation pattern in resting-State default mode network connectivity in firstepisode, treatment-naive major depression patients. Biol. Psychiatry 71, 611–617. Zuo, X., Kelly, C., Adelstein, J.S., Klein, D.F., Castellanos, F.X., Milham, M.P., 2010. Reliable intrinsic connectivity networks: test–retest evaluation using ICA and dual regression approach. NeuroImage 49, 2163–2177.