Abstract

Many psychiatric disorders are associated with abnormal resting-state functional connectivity between pairs of brain regions, although it remains unclear whether the fault resides within the pair of regions themselves or other regions connected to them. Identifying the source of dysfunction is crucial for understanding the etiology of different disorders. Using pathway- and network-based techniques to analyze resting-state functional magnetic imaging data from a large population of patients with attention deficit hyperactivity disorder (239 patients, 251 controls), major depression (39 patients, 37 controls), and schizophrenia (69 patients, 62 controls), we show for the first time that only network-based cross-correlation identifies significant functional connectivity changes in all 3 disorders which survive correction. This demonstrates that the primary source of dysfunction resides not in the regional pairs themselves but in their external connections. Combining pathway and network-based functional-connectivity analysis, we established that, in all 3 disorders, the counterparts of pairs of regions in the opposite hemisphere contribute 60–76% to altered functional connectivity, compared with only 17–21% from the regions themselves. Thus, a transdiagnostic feature is of abnormal functional connectivity between brain regions produced via their contralateral counterparts. Our results demonstrate an important role for contralateral counterpart regions in contributing to altered regional connectivity in psychiatric disorders.

Introduction

An increasing number of functional connectivity studies using resting-state or task-related functional magnetic resonance imaging (fMRI) data from the brain have identified changes in the strength of coupling between pairs of connected regions associated with many mental disorders, including Alzheimer's disease (Wang et al. 2007; Sheline and Raichle 2013), depression (Fitzgerald et al. 2008; Lui et al. 2011; Tao et al. 2013), anxiety (Sylvester et al. 2012), schizophrenia (Liu et al. 2008; Camchong et al. 2011; Guo et al. 2014), autism (Müller et al. 2011), and attention deficit hyperactivity disorder (ADHD) (Mazaheri et al. 2010; Castellanos and Proal 2012). Widespread functional connectivity alterations have been identified in these psychiatric disorders, most prominently in the Default Mode Network: ADHD (Castellanos et al. 2008; Uddin et al. 2008; Castellanos and Proal 2012), depression (Greicius et al. 2007; Zeng et al. 2012), and schizophrenia (Garrity et al. 2007; Whitfield-Gabrieli et al. 2009). Other functional brain network alterations have also been found, for example, dorsal attention (Castellanos and Proal 2012) and prefrontal-striatal networks (Castellanos et al. 2006) in ADHD, affective network in depression (Anand et al. 2005; Greicius et al. 2007; Zeng et al. 2012), and extensive dysconnectivity in schizophrenia (Liang et al. 2006; Lynall et al. 2010). There is also growing interest in establishing transdiagnostic approaches aiming to identify common molecular, neural, and behavioral phenotypes in mental disorders (Buckholtz and Meyer-Lindenberg 2012; Robbins et al. 2012; Consortium C-DGotPG 2013).

Functional connectivity is primarily measured by temporal correlation of activities in pairs of brain regions and analyzed using either cross-correlation (Pearson) (Biswal et al. 1995; Friston et al. 1996) or partial correlation (Marrelec et al. 2006, 2009; Tao et al. 2013) techniques. Despite the extensive research carried out on functional connectivity analysis in mental disorders, it is still unclear whether the cause of the reported changes resides within the region pairs themselves, or in other external regions connected to them in the brain network. If we are going to be able to establish the optimal targets for both diagnostic and therapeutic advances the primary sources responsible for observed functional connectivity changes need to be identified.

Cross-correlation analysis has the advantage that it takes into account contributions from all the different regions in a distributed brain network to the correlation observed between a specific pair of regions. However, while it may be a more accurate reflection of the brain as an interconnected network, cross-correlation techniques may be biased by detecting contributions from third-party regions which are not actually structurally connected (Damoiseaux and Greicius 2009; Honey et al. 2009). This has led to the alternative use of partial-correlation techniques designed to consider functional links between pairs of regions in isolation, thereby removing contributions from all other third-party regions in the brain network (Buckholtz and Meyer-Lindenberg 2012; Robbins et al. 2012; Consortium C-DGotPG 2013). This approach is more accurate in identifying functional connections between region pairs which are also structurally connected because third-party influences are excluded (Zhang et al. 2010). However, the disadvantage is that it may not identify changes occurring at a more complex network level.

In the current study, we have therefore used a combination of cross- (Pearson-) and partial-correlation analysis of resting-state fMRI data in individuals with 3 different psychiatric disorders (ADHD, schizophrenia, and major depression) compared with matched healthy controls. Our hypothesis was that if functional connectivity alterations of patients in pairs of regions are primarily contributed to by changes within these regions, then partial correlation, which removes third-party influences and reflects direct interaction between a pair of regions, should be the most sensitive in detecting the functional connectivity alterations. If on the other hand such changes are primarily contributed by third-party regions, then the Pearson-correlation technique should be more sensitive. If the latter proved to be the case, we could also combine these 2 approaches to identify which third-party regions contribute most to observe functional connectivity changes. Here, we adopted a triplets-regions of interest (ROI)-based partial-correlation approach especially suited to identify the influences from each individual third-party mediator. Our work therefore is expected to provide new insights into the mechanisms of how large-scale organization of functional networks changes in the disordered brain and therapeutic strategies.

Materials and Methods

Subjects

Three resting-state datasets were used from ADHD, depression, and schizophrenia patients and their respective healthy control groups. Full demographic details of patients and healthy controls are given in Supplementary Table 1. The ADHD dataset (Database 1) was obtained from the ADHD-200 Consortium for the Global Competition (http://fcon_1000.projects.nitrc.org/indi/adhd200/) and includes a total of 490 subjects (251healthy controls and 239 ADHD patients—55 patients are identified as medicated and 85 as unmedicated. For the remaining 99 patients no information is given). All ADHD patients and healthy controls were evaluated using the Schedule of Affective disorders and schizophrenia for Children—Present and Lifetime version (KSADS-PL) with one parent for the establishment of diagnosis. The ADHD Rating Scale (ADHD-RS) IV was employed to measure severity of ADHD symptoms. All patients and healthy controls were either from China or USA and (1) right handed, (2) no history of head trauma with loss of consciousness (3) no history of neurological disease or diagnosis of schizophrenia, affective disorder, pervasive development disorder, or substance abuse (4) had a Weschler Intelligence Scale for Children score of >80.

The major depression database (Database 2) includes 76 Chinese subjects (37 healthy controls and 39 depression patients—15 first episode depression who were unmedicated and 24 medicated, treatment-resistant depression) from Second Xiangya Hospital, Central South University in Changsha, Hunan Province, China and also reported in a previous paper (Tao et al. 2013). All patients met the following inclusion criteria: (1) current major depressive disorder attack as assessed by 2 experienced psychiatrists using the Structural Clinical Interview for DSM-IV; (2) 18–45 years of age; (3) right-handed Han Chinese; (4) Hamilton Rating Scale for depression scores of at least 17; (5) treatment-naive adult patients with first episode major depression had not taken any medication before the MRI scan. Patients and healthy controls met the following exclusion criteria: (1) a history of neurological diseases or other serious physical diseases; (2) a history of electroconvulsive therapy; (3) a history of substance abuse (4) comorbidities with other disorders (no evidence for schizoaffective disorder or Axis II, personality disorders and mental retardation).

The schizophrenia dataset (Database 3) includes 131 Taiwanese subjects (62 healthy controls and 69 medicated schizophrenia patients) from National Taiwan University Hospital in Taiwan and reported in a previous paper (Guo et al. 2014). All patients were identified according to DSM-IV diagnostic criteria by qualified psychiatrists and symptom severity assessed using the Positive and Negative Syndrome Scale (PANSS). Exclusion criteria included (1) the presence of other DSM-IV disorders; (2) history of substance abuse; (3) clinically significant head trauma. Healthy controls were also confirmed using DSM-IV criteria to be free of schizophrenia or other Axis 1 disorders and not to have a history of substance abuse or clinically significant head trauma. All patients and healthy controls were right handed with the exception of 2 patients who were left handed.

Details of drug treatments for patients in the 3 datasets are provided in Supplementary Table 2. Relevant ethical permissions for experiments and subject individual consent forms were obtained for all 3 datasets (Dataset 1 can be found at http://fcon_1000.projects.nitrc.org/indi/adhd200/; Dataset 2 is depression data (Tao et al. 2013), and Dataset 3 is schizophrenia data (Guo et al. 2014)).

Image Acquisitions and Data Preprocessing

A strict criterion of head movement not greater than ±1.5 mm and ±1.5° was used. For Dataset 1 (ADHD) resting-state functional imaging data were acquired from Peking University and New York University.

Peking University All functional imaging data and T1-weighted images were acquired using a SIEMENS TRIO 3 T. A black screen with a white fixation cross was displayed during the scan. Participants were asked to relax and keep still when functional imaging data were collected. A total of 232 volumes of echo planar images (EPI) were obtained axially (slices, 30; repetition time (TR), 2000 ms; echo time (TE), 30 ms; slice thickness, 4.5 mm; flip angle 90°; field of view (FOV), 220 × 220 mm2; matrix, 64 × 64). A high-resolution T1-weighted magnetization prepared rapid acquisition gradient echo (mprage) image was obtained for all subjects, see Supplementary Table 1(a) for details.

New York University Participants were asked to remain still, close their eyes and think of nothing systematically but not to fall asleep when functional imaging data were collected. A black screen was presented to them. Image data were acquired using Siemens Allegra 3.0 T scanner. A total of 172 volumes of EPI images were obtained axially (slices, 33; TR, 2000 ms; TE, 15 ms; slice thickness, 4 mm; flip angle, 90°; FOV, 240 × 240 mm2; matrix, 80 × 80). A high-resolution T1-weighted mprage was obtained for each subject (slices, 128; TR, 2530 ms; TE, 3.25 ms; inversion time, 1100 ms; slice thickness, 1.33 mm; flip angle, 7°; FOV, 256 × 256; matrix, 256 × 256).

For Dataset 2 (depression), image data were acquired using a 1.5 T Siemens MRI scanner. Individuals were instructed to keep their eyes closed but not go to sleep. A total of 180 volumes of EPI images were obtained axially (slices, 20; TR, 2000 ms; TE, 40 ms; thickness, 5 mm; gap, 1 mm; flip angle, 90°; FOV, 240 × 240 mm2; resolution, 64 × 64). High-resolution whole-brain T1-weighted images were acquired sagittally with a 3D spoiled gradient echo pulse sequence (TR, 12.1 ms; TE, 4.2 ms; flip angle, 151; FOV = 240 × 240 mm2; acquisition matrix, 256 × 256; thickness, 1.8 mm; number of excitations, 2; 172 slices).

For Dataset 3 (schizophrenia), all subjects underwent a structural and functional MRI scan in a single session using a 3T MR system (TIM Trio, Siemens). Individuals were instructed to keep their eyes closed but not go to sleep. A total of 180 volumes of EPI images were obtained axially, (slices, 34; TR, 2000 ms; TR, 24 ms; thickness, 3 mm; flip angle, 90°; FOV, 256 × 256 mm2; resolution, 64 × 64). A whole-brain high-resolution T1-weighted MR image was acquired using a magnetization prepared rapid gradient echo (MPRAGE) sequence in a coronal view (TR, 2000 ms; TE, 2.98 ms; inversion time, 900 ms; image matrix size, 192 × 256, spatial resolution, 1 × 1 mm2, FOV, 192 × 256 mm2; slice thickness, 1 mm).

For the 3 datasets prior to preprocessing, the first 10 volumes were discarded to allow for scanner stabilization and the subjects' adaptation to the environment. fMRI data preprocessing was then conducted by SPM8 (http://www.fil.ion.ucl.ac.uk/spm) and a Data Processing Assistant for Resting-State fMRI (DPARSF). The remaining functional scans were first corrected for within-scan acquisition time differences between slices and then realigned to the middle volume to correct for interscan head motions. Subsequently, the functional scans were spatially normalized to a standard template (Montreal Neurological Institute) and resampled to 3 × 3 × 3 mm3. After normalization, blood oxygen level-dependent (BOLD) signal of each voxel was firstly detrended to abandon linear trend and then passed through a band-pass filter (0.01–0.08 Hz) to reduce low-frequency drift and high-frequency physiological noise. Finally, nuisance covariates including head motion parameters, global mean signals, white matter signals, and cerebrospinal signals were regressed out from the BOLD signals. After data preprocessing, the time series were extracted in each ROI by averaging the signals of all voxels within that region.

For all 3 datasets, the automated anatomical labeling atlas (Tzourio-Mazoyer et al. 2002) was used to parcellate the brain into 90 ROIs (45 per hemisphere). The names of the ROIs and their corresponding abbreviations are listed in Supplementary Table 3.

Functional Connectivity Analysis

The BOLD signals of all voxels were obtained by a band-pass filter, extracting the low frequencies of interest (0.01–0.08 Hz). Regional BOLD signals for each individual were obtained by averaging the time series of all the voxels in the region. For all subjects, both cross-correlation (Pearson-correlation) and partial-correlation analysis were performed to measure whole-brain functional connectivity using regional BOLD time series. We used a triplets-ROI-based partial-correlation approach to remove the mediation from third-party regions. For an arbitrary pair of regions i and j, its partial correlation is calculated multiple times, each time with one third-party region k (k = 1, 2, … ,90, ki and kj) being controlled (i.e., the mediation from region k is removed). We call these 3 regions i, j, and k a triplet in this case. Since there are altogether 90 brain regions, there will be 88 third-party mediators for i and j (i.e., 88 triplets), and thus 88 partial-correlation coefficients will be obtained for region pair i and j. We then pick the smallest one (in absolute value) as the partial-correlation coefficient between i and j (Tao et al. 2013; Guo et al. 2014), indicating that the largest influence among all third-party mediators is removed. The corresponding third-party region, which has the largest mediation effect and leads to the smallest partial correlation (in absolute value), is called the primary mediating region.

The reason why we used this triplets-based approach to estimate partial correlation is 2-fold: (1) the number of samples (length of BOLD signal) is small compared with the number of variables (brain regions) that traditional inverse covariance matrix approach may lead to poor estimate; our approach avoids this problem by performing a first-order estimate, that is, removes the mediation from one third-party region at a time, and then pick the smallest partial correlation (i.e., removing the largest mediation); (2) Our approach allows evaluation of the mediating strength from each individual third-party region to a given functional pair (see below), therefore, we can spot the third-party mediator that exerts the largest influence. The goal of partialling out the influence from one third-party mediator at a time is not to discard useful information, but to identify which region has the greatest influence to a given pair of regions.

After obtaining the whole-brain functional connectivity using both Pearson- and partial-correlation, we then performed 2-sample t-test between patients and matched controls for each of the functional connectivity. The difference across the 2 groups was only considered significant where they survived false discovery rate (FDR) (q < 0.025). We used a stringent correction here to avoid type-I error in multiple comparison. For schizophrenia data, we used a strict Bonferroni correction as there would otherwise be too many functional connectivity changes (by Pearson correlation) identified using a looser correction. Finally, to evaluate the correlation between altered functional connectivity in patients and corresponding symptom scores of various diseases, Pearson correlation analysis is performed.

Identifying Primary Mediating Regions and the Strength of Their Influence

For a given pair of brain regions i and j, the above triplets-ROI-based partial-correlation analysis allows us to evaluate the mediation, or contributions from each individual third-party region. Pearson-correlation between i and j (denoted by Pe) embraces simultaneously mediations from all third-party regions, while partial correlation with region k being the controlled variable (denoted by Pak) reflects functional connectivity between i and j after removing the mediation of region k. Therefore, the mediation, or influence exerted by k to the functional connectivity between i and j can be defined as Pe Pak.

It should be noted that there are 88 third-party mediators for each region pair i and j, thereby forming 88 triplets, each of which generates a partial-correlation coefficient. We have chosen the smallest one (in absolute value) as the final partial correlation between i and j, because the largest influence among all third-party regions is removed in this way. We call this third-party region that exerts the largest influence to the given pair i and j as the primary mediating region (denoted by K), with its mediating strength being Pe PaK. Similarly, if we choose the third-party region that lead to the second smallest partial correlation (among all 88 partial correlations), then this third-party region is defined as the secondary mediating region, with the corresponding mediating strength being the secondary mediating strength. Finally, a positive mediation indicates that third-party region influences i and j in the same direction, for example, excites (or inhibit) both of them. A negative mediation means the third-party region is affecting i and j in different directions.

Support Vector Machine (SVM) Classifier

The SVM is a learning machine for a 2-class classification problem widely used because of its ability to handle very high-dimensional data and due to its accuracy in classification and prediction. In the current study we used the same method (Guo et al. 2013) to calculate the prediction accuracy for the altered functional connections identified by cross-correlation and also the contralateral mediating strength in the 3 different disorders. We used SVM toolkit libsvm (http://www.csie.ntu.edu.tw/~cjlin/libsvm/, by Lin Chih-Jen of Taiwan University). A radial basis function is selected as the kernel function (t = 2) and parameter C is fixed to 10 to trade-off learning and generalization while other parameters are kept as default values. Statistical significance of the accuracy estimates was also calculated using a permutation analysis.

Results

Functional Connectivity Changes in ADHD, depression, and schizophrenia

We first carried out a whole-brain functional-connectivity analysis comparing patient groups and their respective healthy controls using both Pearson- and partial-correlation analysis. Results revealed that only Pearson-correlation identified significantly altered pathways in all 3 disorders following standard FDR correction (see Supplementary Table 3 for abbreviations of brain regions. ADHD: 50 connections, depression: 18 connections, and schizophrenia: 70 connections see Fig. 1 and Supplementary Tables 4–6). Partial correlation failed to identify any changes that survived FDR correction. This failure of partial correlation therefore showed that the significant functional connectivity changes found using Pearson-correlation must primarily have been contributed to by third-party regions. This is further emphasized by the finding that partial correlation for the same functional connections identified by Pearson as significantly altered in the 3 disorders were considerably lower than those calculated using Pearson (see Fig. 1 and Supplementary Tables 4–6).

Figure 1.

Regional links showing significantly altered functional connectivity identified by Pearson correlation in (a) ADHD, (b) depression and (c) schizophrenia. Corresponding alterations in partial correlation are also given for each link although none of these are significant (for abbreviations see Supplementary Table 3). “*” denotes that the altered connectivity is interhemispheric. Where regions are underlined this indicates that their contralateral counterparts are first or second primary mediators of the corresponding altered link.

Figure 1.

Regional links showing significantly altered functional connectivity identified by Pearson correlation in (a) ADHD, (b) depression and (c) schizophrenia. Corresponding alterations in partial correlation are also given for each link although none of these are significant (for abbreviations see Supplementary Table 3). “*” denotes that the altered connectivity is interhemispheric. Where regions are underlined this indicates that their contralateral counterparts are first or second primary mediators of the corresponding altered link.

In ADHD significantly altered connections were found in a range of frontal, parietal, temporal, occipital, subcortical, and limbic areas with the most affected regions being the orbitofrontal cortex, inferior and superior frontal gyrus, anterior and posterior cingulate gyrus, calcarine cortex, and parahippocampal gyrus. For depression changes were primarily in orbitofrontal cortex, posterior cingulate cortex, parietal and temporal cortices, hippocampus, amygdala, insula, caudate, putamen, pallidum and thalamus. Schizophrenia patients showed the most widespread changes involving all major cortical and subcortical subdivisions. However, the most affected regions were medial and superior frontal gyri, medial and posterior cingulate gyri, pre- and postcentral gyri, angular gyrus, fusiform and lingual gyri, thalamus, and caudate. In all 3 disorders approximately half of the altered functional connections were within brain hemispheres and half between them (within vs. between hemispheres: ADHD = 24/26; depression = 10/8; schizophrenia: 34/36 links).

Which Third-party Regions are Responsible for Altered Functional Connectivity?

Since partial correlation between a pair of brain regions (which removes mediations from third-party regions) is not sensitive in finding significant functional-connectivity changes, while Pearson-correlation (which contains third-party mediation) can detect the changes, this suggests that the third-party mediation is vital for the functional-connectivity alteration in patients. By triplets-ROI-based partial-correlation analysis, we manage to find which third-party region played the most significant role in contributing to altered functional connections between pairs of brain regions in ADHD, depression, and schizophrenia. This revealed that the main contributors were the 2 contralateral counterpart regions for many of these regional pairs exhibiting altered functional connectivity (Fig. 2a and Supplementary Tables 7–9). We defined contralateral mediating strength as the larger one (in absolute value) contributed to by the 2 individual counterpart regions. Overall, we found that the change in contralateral mediating strength (across control and patient group) contributed 60–76% of the altered Pearson-correlation found in patients in individual regional pairs, whereas the change within the pairs themselves, that is, partial correlation, accounted for only 17.5–21% (Table 1). This finding indicates that it is altered mediation by contralateral counterpart regions which is primarily responsible for the changes in functional connectivity observed in regional pairs in patients in all 3 disorders. Note here “change” means the difference between control and patient groups.

Table 1

The mean change (across controls and patients) in functional connectivity measured by partial correlation, the contralateral mediating strength, and Pearson correlation for all the altered functional connections in ADHD (50 links), depression (18 links), and schizophrenia (70 links), respectively

Disorder Partial-correlation change Contralateral counterpart mediation change Pearson correlation change 
(a) Using the original correlation value 
 ADHD 0.0127 (17.5%) 0.0466 (64.3%) 0.0725 (100%) 
 Depression 0.0413 (21.0%) 0.1180 (60.1%) 0.1965 (100%) 
 Schizophrenia 0.0336 (18.5%) 0.1706 (76.2%) 0.2238 (100%) 
(b) Using the square of the original correlation value 
 ADHD 0.0016 (6.1%) 0.0137 (52.3%) 0.0262 (100%) 
 Depression 0.0035 (6.7%) 0.0253 (48.8%) 0.0518 (100%) 
 Schizophrenia 0.0033 (5.7%) 0.0426 (73.3%) 0.0581 (100%) 
Disorder Partial-correlation change Contralateral counterpart mediation change Pearson correlation change 
(a) Using the original correlation value 
 ADHD 0.0127 (17.5%) 0.0466 (64.3%) 0.0725 (100%) 
 Depression 0.0413 (21.0%) 0.1180 (60.1%) 0.1965 (100%) 
 Schizophrenia 0.0336 (18.5%) 0.1706 (76.2%) 0.2238 (100%) 
(b) Using the square of the original correlation value 
 ADHD 0.0016 (6.1%) 0.0137 (52.3%) 0.0262 (100%) 
 Depression 0.0035 (6.7%) 0.0253 (48.8%) 0.0518 (100%) 
 Schizophrenia 0.0033 (5.7%) 0.0426 (73.3%) 0.0581 (100%) 

The change in Pearson correlation between a given pair of regions is contributed by change in direct interaction (i.e., partial correlation) between this pair of regions and the change of mediation from third-party regions (e.g., contralateral counterparts to the given pair of regions), thus that contributed by the contralateral counterpart mediation (60–76%). Furthermore, when added together Pearson correlation change can be deemed as overall change (i.e., 100%). It can be seen in Table 1 (a) that the actual change in direct interaction (measured by partial correlation) is much smaller (17.5–21%) than the 2 contributions from the functional link itself and its primary contralateral mediator account for 81–95% of the total change measured by Pearson correlation. Table 1 (b) is the same as Table 1 (a), but is obtained using the square of the original correlation value. Table 1 (b) shows that the contribution from contralateral counterpart mediation is much larger than that from the direct interaction of a functional pair (0.0137/0.0016 = 8.6 times for ADHD, 0.0253/0.0035 = 7.3 times for depression, and 0.0426/0.0033 = 12.9 times for schizophrenia).

Figure 2.

Schematic drawing of contralateral mediation for a specific functional connectivity. The top and bottom panel are for interhemispheric (a–c) and intrahemispheric connectivity (d–f), respectively. In the top panel, the functional connectivity between xR and y (black circles) is mediated by the contralateral symmetric region of xR, that is, xL (green circle). (a) shows a strong functional connectivity (dark red line) between xR and y calculated using Pearson correlation because it is contributed to by the direct interaction between xR and y (partial correlation, pink line, which excludes contributions from xL, see b), and by contralateral mediation from region xL on both xR and y (red line, see c). This is the same for the intrahemispheric connectivity in bottom panel, that is, the strong functional connectivity between xL and y calculated using Pearson correlation (in d) is contributed to by the direct interaction between xL and y (partial correlation, which excludes contributions from xR, see e), and by contralateral mediation from region xR on both xL and y (see f). We found that for the altered functional connectivity of all 3 disorders, there is always motif-3 structure whereby both the right (xR) and left sides (xL) of the same region showed altered functional connectivity with another region (y) such that for xR to y, xL is the primary mediator and for xL to y, xR is the primary mediator.

Figure 2.

Schematic drawing of contralateral mediation for a specific functional connectivity. The top and bottom panel are for interhemispheric (a–c) and intrahemispheric connectivity (d–f), respectively. In the top panel, the functional connectivity between xR and y (black circles) is mediated by the contralateral symmetric region of xR, that is, xL (green circle). (a) shows a strong functional connectivity (dark red line) between xR and y calculated using Pearson correlation because it is contributed to by the direct interaction between xR and y (partial correlation, pink line, which excludes contributions from xL, see b), and by contralateral mediation from region xL on both xR and y (red line, see c). This is the same for the intrahemispheric connectivity in bottom panel, that is, the strong functional connectivity between xL and y calculated using Pearson correlation (in d) is contributed to by the direct interaction between xL and y (partial correlation, which excludes contributions from xR, see e), and by contralateral mediation from region xR on both xL and y (see f). We found that for the altered functional connectivity of all 3 disorders, there is always motif-3 structure whereby both the right (xR) and left sides (xL) of the same region showed altered functional connectivity with another region (y) such that for xR to y, xL is the primary mediator and for xL to y, xR is the primary mediator.

The proportion of the altered connections in the 3 disorders where a contralateral counterpart was either the first (primary) or second (secondary) strongest mediating region was also very high (ADHD—88%; depression—94.4%; schizophrenia—80%—see Supplementary Tables 7–9). Indeed, overall there was a significant positive correlation between altered functional connectivity in region pairs in all 3 disorders and altered mediating strength from the most influential contralateral counterpart region (see Fig. 3). A further analysis of the proportion of altered functional connections where the mediating strength of contralateral counterpart regions was actually significantly (P < 0.05) different between patients and controls showed this to be 35/44 (79.5%) in ADHD, 11/17 (64.7%) in depression and 54/56 (96.4%) in schizophrenia with overall levels of significance being higher for changes in schizophrenia. In all 3 disorders the majority of these mediation influences were weakened (ADHD—42%; depression –55.6%; schizophrenia—45.7%) with only a small proportion being strengthened (ADHD—18%; depression—5.6%; schizophrenia—7.1%—see Supplementary Table 10).

Figure 3.

Significant positive correlations between the difference of the magnitude of altered functional connectivity (Pearson) across controls and patients and that of associated contralateral mediating strength for (a) ADHD (50 links), (b) depression (18 links) and (c) schizophrenia (70 links) patients. In all cases absolute values and the contralateral counterpart region with the strongest mediating strength were used.

Figure 3.

Significant positive correlations between the difference of the magnitude of altered functional connectivity (Pearson) across controls and patients and that of associated contralateral mediating strength for (a) ADHD (50 links), (b) depression (18 links) and (c) schizophrenia (70 links) patients. In all cases absolute values and the contralateral counterpart region with the strongest mediating strength were used.

A number of regions with altered functional links in the 3 disorders were also identified as significant mediators (see Supplementary Tables 7–9). A notable motif revealed by our analysis is that in many cases a 3-region bilateral interactive network is formed whereby if the right side of a region (xR) shows altered connectivity with another one (y) in patients, then the left side (xL) of the same region acts as the key mediating structure (see Fig. 2a-c, in which xL serves as the mediator for link xR to y). Additionally, the link between xL and y is also often altered, and in this case xR acts as the key mediating structure (see Fig. 2d–e). This pattern whereby xR and xL both act as mediators for their counterpart's altered link with y occurs in a number of cases in all 3 disorders (see Fig. 4). The most marked in ADHD is for the right and left medial orbitofrontal cortices which have this motif with the right and left parahippocampal gyrus, right and left calcarine cortex, and left rectus gyrus (i.e., 14 altered links); in depression the right and left putamen, right and left pallidum, and right and left insula all have this motif in altered links with the right inferior orbitofrontal cortex (i.e., 6 altered links); in schizophrenia the right and left thalamus had this motif with the left and right medial frontal gyrus, left and right postcentral gyrus, left and right fusiform gyrus, right precentral gyrus, and left lingual gyrus (i.e., 18 altered links). Furthermore, the sign of the difference between functional connectivity strengths in patients versus controls for xR to xL, xR to y, and xL to y is the same (i.e., always positive or always negative) in almost all cases, indicating potential additive effects involving all 3 links. Overall functional connectivity strengths between bilateral counterpart regions (i.e., xR to xL) were always higher than for any other connections (see Supplementary Table 11).

Figure 4.

Functional connections and their associated mediating regions significantly correlated with symptom scores. (a) ADHD, (b) depression (only main altered connections are shown since there are no correlations with symptoms), (c) schizophrenia (positive-symptom-related connections), (d) schizophrenia (negative-symptom-related connections). Functional connections (solid lines) and mediating links (dashed lines) are either reduced in strength in patients compared with healthy controls (blue) or increased (red). Connections are either positively correlated with symptom severity (dark red/blue) or negatively correlated (light red/blue). In all cases the direction of the changes in functional connectivity and that of mediating strength is the same. Regions only associated with functional connectivity changes are denoted by large nodes (black) and where they are only mediators as smaller nodes (green). Where regions are involved in both functional connectivity and mediating then a large node includes a smaller one inside it. In all cases the mediation influence on a specific functional connection is on both regions involved and only the contralateral counterpart region with the strongest mediating influence is shown.

Figure 4.

Functional connections and their associated mediating regions significantly correlated with symptom scores. (a) ADHD, (b) depression (only main altered connections are shown since there are no correlations with symptoms), (c) schizophrenia (positive-symptom-related connections), (d) schizophrenia (negative-symptom-related connections). Functional connections (solid lines) and mediating links (dashed lines) are either reduced in strength in patients compared with healthy controls (blue) or increased (red). Connections are either positively correlated with symptom severity (dark red/blue) or negatively correlated (light red/blue). In all cases the direction of the changes in functional connectivity and that of mediating strength is the same. Regions only associated with functional connectivity changes are denoted by large nodes (black) and where they are only mediators as smaller nodes (green). Where regions are involved in both functional connectivity and mediating then a large node includes a smaller one inside it. In all cases the mediation influence on a specific functional connection is on both regions involved and only the contralateral counterpart region with the strongest mediating influence is shown.

We also carried out a separate analysis to determine if the general pattern of weakened mediating effects by contralateral counterpart regions in the 3 disorders might have resulted from the medications used. Medicated and unmedicated ADHD or depression patients were compared separately with their respective control groups. While contralateral mediation rate is higher in medicated patients than unmedicated patients (which might be due to medication effect), in both cases there was still an overall reduction in mediating strength (see Supplementary Table 12). Thus, it would appear unlikely that our findings were influenced by nonspecific medication effects.

Associations between Functional Connectivity and Contralateral Mediation Changes and Symptom Severity/Illness Duration

Functional connectivity (Pearson correlation) changes significantly associated with symptom severity in the 3 disorders are shown in Figure 4 and Supplementary Tables 13–15. For ADHD, 7/50 functional pathways were associated with symptom severity. These included the right medial orbitofrontal cortex connections with left middle orbitofrontal cortex and right inferior frontal gyrus (triangular) (positive correlation) and the left medial orbitofrontal cortex with the left and right calcarine cortex (negative correlation); the right and left inferior frontal gyrus (triangular) connections to the left rectus gyrus and right anterior cingulate gyrus, respectively (positive correlation) and left medial frontal gyrus connection with right inferior temporal gyrus (negative correlation, see Fig. 4a). For depression patients we did not find any significant correlations with Hamilton scores, although these were mostly high and had a narrow range in our patient groups. Figure 4b shows the 11 out of 18 altered links showing contralateral mediation and including the 2 most frequently involved regions, the medial orbitofrontal cortex (8 links) and inferior parietal lobule (3 links). For schizophrenia 22/70 functional connections were found to correlate significantly with PANSS scores (13 with positive and 9 with negative symptom severity, see Fig. 4c,d). For positive symptoms connections involved either thalamo-frontal (6 links—negative correlation) or thalamo-postcentral gyrus/rolandic operculum (3 links—positive correlation) links; medial cingulate gyrus to superior occipital gyrus (negative correlation); caudate to middle temporal pole (2 links—positive correlation) and right posterior cingulate cortex to left medial frontal gyrus (negative correlation) and left fusiform gyrus (positive correlation). For negative symptoms connections involved either the posterior cingulate gyrus to fusiform and lingual gyri (5 links—positive correlation) and medial cingulate gyrus to pallidum and putamen (4 links—negative correlation). In addition, 13/70 functional pathways were significantly associated with illness duration in schizophrenia. These had a large overlap (6/13 links) with functional connections associated with negative symptom severity and included medial cingulate gyrus to putamen, pallidum, and amygdala (6 links—negative correlation) and posterior cingulate gyrus to lingual gyrus (3 links—positive correlation). Other links associated with illness duration included left lingual gyrus to bilateral amygdala (positive correlation), left to right pallidum (negative correlation) and left angular gyrus to left superior occipital gyrus (positive correlation) (see Supplementary Table 16).

A further correlation analysis between strengths of the 2 contralateral mediating regions influencing each of the altered functional pathways and symptom severity revealed that in the majority of cases they were also significantly correlated with symptom severity (ADHD: 8 and 7 links, see Supplementary Table 17a and b; schizophrenia—positive symptoms: 15 and 20 links, see Supplementary Table 18a and b; negative symptoms: 9 and 10 links—see Supplementary Table 19a and b). In schizophrenia both contralateral counterpart mediating regions were often correlated with symptom severity (5 links for positive symptoms and 5 for negative symptoms). For illness duration in schizophrenia there was a similar pattern with 10/13 of the links showing significant associations between the altered mediating strength from either of the 2 contralateral regions (see Supplementary Table 20). In all cases the direction of the correlation (positive or negative) was the same for altered functional links and altered contralateral region mediating strength.

The regions showing significantly altered functional connectivity and contralateral counterpart mediation associated with symptom severity in ADHD and for schizophrenia are shown in Figure 4, Supplementary Tables 13–15 and 17–19. It can be seen that for ADHD the main links are those involving the medial orbitofrontal cortex and calcarine cortex. For schizophrenia, links associated with positive symptoms include more frontal (orbitofrontal and medial frontal) and medial (medial cingulate, caudate and thalamus) structures whereas for negative symptoms medial (medial cingulate gyrus and pallidum) and posterior (posterior cingulate gyrus and lingual gyrus) structures are. For illness duration associations were also with medial and posterior structures (see Supplementary Tables 16 and 20).

SVM Analysis of the Predictive Value of Functional Connectivity Changes

Classification accuracy revealed by support vector machine (SVM) showed that pathways showing altered functional connectivity identified by Pearson-correlation were effective for distinguishing patients from healthy controls (leave-one-out accuracy: ADHD: 68.0%; depression: 85.5%; schizophrenia: 84.7%, respectively, all P < 0.001). A separate analysis of the accuracy of altered mediating strength in contralateral counterpart regions for discriminating patients and controls in the 3 disorders revealed similar results (ADHD: 64.7%; depression: 72.4%, schizophrenia: 82.4%, all P < 0.001, see Supplementary Table 21 for details. We did not use partial correlation in the SVM analysis as it failed to identify any changes that survived FDR correction.

Discussion

Overall we have provided the first systematic investigation of what contributes to altered functional coupling between region pairs in the disordered brain through a combined use of cross-correlation and partial-correlation techniques. It is noteworthy that the goal of partialling out third-party mediation (in our triplets-ROI-based partial-correlation analysis) is not to discard information arbitrarily but, on the contrary, to evaluate the third-party mediation and, to identify the mediator exerting the greatest mediation to a given pair of regions. This revealed that significant changes observed in patients are not due primarily to the specific region pairs themselves but to altered mediation influences via third-party structures. In 80–94% of links in ADHD, depression and schizophrenia patients the first or second most influential mediator of altered functional connectivity was one of the 2 contralateral counterpart regions, and in a high proportion of these the change in mediating strength in patients was also significant, most notably in schizophrenia (93%). The overall contribution from these counterpart regions accounted for 60–76% of the functional-connectivity change between region pairs in patients compared with controls, whereas changes restricted to the region pairs themselves only accounted for 17.5–21%. Furthermore, associations between symptom severity and illness duration and altered functional connectivity found in region pairs were reflected in most cases by similar alterations in the mediating strength from their contralateral counterparts. A SVM analysis showed good discrimination of patients and healthy controls either using Pearson correlation (68–86%) or contralateral mediating strength (65–82%). Our results illustrate both that more attention should be paid in future towards the contribution of these contralateral mediating regions rather than specific altered links in psychiatric and other mental disorders and that a common motif in the disordered brain may be altered interhemispheric communication/integration.

The Importance of Altered Interhemispheric Communication in Mental Disorders

There is considerable evidence for altered functional and/or structural hemispheric connectivity/communication in a number of mental disorders including Alzheimer's and mild cognitive impairment (Di Paola et al. 2010), depression (Xu et al. 2013), anxiety (Compton et al. 2008), bipolar disorder (Bearden et al. 2011), schizophrenia (Crow 1998; Knöchel et al. 2012; Guo et al. 2013, 2014), ADHD (Gilliam et al. 2011), Tourette's (Plessen et al. 2004), autism (Anderson et al. 2011), borderline personality disorder (Rüsch et al. 2010), and disorders of consciousness (Ovadia-Caro et al. 2012). Reduced interhemispheric communication resulting from congenital agenesis or sectioning of the corpus callosum has also been associated with impaired cognitive and emotional functioning (Bloom and Hynd 2005; van der Knaap and van der Ham 2011). Around 50% of altered functional connections in each of the 3 disorders in the current study involved different regions in the 2 hemispheres, and this together with the high proportion of altered links with significantly altered mediation via their contralateral counterparts indicates that there is a contribution from both hemispheres to dysfunction in 120 out of 138 links (87%). Further, in 100 out of these 120 links there are significantly altered mediating strength from a contralateral counterpart region and which mainly reflects a weakened influence.

In support of previous studies (Tao et al. 2013) we found that functional connectivity between counterpart regions in the 2 hemispheres is generally very strong and reflects extensive fiber links between bilateral structures involving the corpus callosum or commissures (van der Knaap and van der Ham 2011). As we used a rather stringent correction for multiple comparison, alteration of functional connectivity between bilateral regions themselves in the 2 hemispheres is not very significant (only 1/138 altered links was bilateral). However, using the same dataset we have shown recently that in both schizophrenia and depression there is a general overall reduction in functional connectivity involving them (Guo et al. 2013). Furthermore, although changes in functional connectivity between bilateral pairs of regions did not achieve significance the absolute magnitude of the differences between controls and patients was often similar or even larger than that observed in other less strongly connected links which did achieve significance. Thus it is possible that the very high and more variable functional correlation strengths found between bilateral structures may have contributed to observed changes failing to achieve significance.

Finally, it is important to note that our findings that region xR being identified in many cases as the primary mediating region for functional connectivity between xL and y is not simply due to the generally high correlation between homologous pair of regions xR and xL. In order for the above findings to hold, region xR must simultaneously have strong functional connectivity with region y. In other words, the above findings are not simply the result of the strong functional connectivity between bilateral regions xR and xL; they only occur when xR influences, or mediates both xL and y in a significant way. Furthermore, it is just the changes in this contralateral region mediation that lead to the significantly altered functional connectivity in patient group, see Supplementary Tables 7–9.

How do Contralateral Counterpart Regions Influence Altered Functional Connectivity?

The influence of contralateral mediating structures could be both direct and indirect via other third-party regions. It is possible that the altered mediation is via the direct connections with contralateral counterparts since, as discussed above, although these do not exhibit significant changes in functional connectivity themselves in the disorders they are the strongest functional connections. Thus even very small changes in their mediation influence might produce large effects. On the other hand, we have shown that in many cases there is a 3-motif relationship whereby both the right (xR) and left sides (xL) of the same region showed altered functional connectivity with another region (y), such that for xR to y, xL is the mediator and for xL to y, xR is the mediator (see Fig. 2). Thus, xR and xL may also exert their mediation effects indirectly via y. With the methodology used in the current paper, we cannot distinguish between these 2 potential routes of mediation effects.

Finally, it should be noted that the term “mediation”, or “influence” is from a triplets-based partial-correlation point of view, that is, if the functional connectivity between a pair of regions is significantly changed by partialling out a third-party region, we believe that this third-party region will have a mediation effect influencing a given pair. In this case Granger Causality Analysis needs to be conducted in the future to clarify the causal relation among these triplets ROIs (xR and xL and y), and identify the possible mediation pathways.

What are the Main Functional Connectivity Changes in ADHD, depression, and Schizophrenia?

Although we found 50 altered functional connections in ADHD, 18 in depression, and 70 in schizophrenia we will focus our discussion mainly on groups of connections where right and left regions of some key structures play reciprocal roles as mediators and are interconnected with a widespread network. The relevance of these particular altered circuits is also underlined by the fact that they account for the majority of links associated with symptom severity in both ADHD and schizophrenia.

In ADHD the most notable altered links exhibiting the above criteria involved the medial orbitofrontal cortex and its functional connections with the calcarine cortex, although many altered functional connections with other frontal regions (middle orbitofrontal cortex, inferior frontal gyrus, rectus gyrus, and anterior cingulate gyrus) were found. There is increasing evidence for altered resting-state and task-related connectivity between frontal cortex and primary visual cortex in ADHD (Mazaheri et al. 2010; Castellanos and Proal 2012). The medial orbitofrontal cortex receives inputs from the ventral visual processing stream involved in object recognition, and plays important roles in control of impulsivity and reward (Rolls 2004). Impaired impulsivity is a key symptom in ADHD (Robbins et al. 2012) and is also associated with orbitofrontal cortex volume changes (Hesslinger et al. 2002; Boes et al. 2009). Thus, reduced functional connectivity between medial orbitofrontal and visual cortices suggests a reduced ability of visual cues from social or other stimuli to elicit appropriate impulse control or reward. Previous studies have also emphasized fronto-striatal disconnection in ADHD (Castellanos and Proal 2012). Functional connectivity between the thalamus and putamen showed significantly increased connectivity in patients but this had no association with symptom severity. This supports previous findings in ADHD relating changes in this functional link to impaired spatial working memory capacity (Mills et al. 2012).

In depression, the right inferior orbitofrontal cortex is involved in 8/18 altered functional connections including reduced connectivity with the bilateral putamen, insula, and pallidum and left hippocampus and thalamus. While the left inferior orbitofrontal cortex has no altered functional links it exhibits significantly reduced mediating strength in relation to right orbitofrontal cortex functional connections with the bilateral pallidum and left insula. Our previous studies have also found evidence for both reduced gray matter volume (Scheuerecker et al. 2010) and either increased or decreased functional connectivity (Frodl et al. 2010; Lui et al. 2011; Tao et al. 2013) involving the orbitofrontal cortex in major depression which may reflect altered responses to rewarding stimuli and anhedonia (Treadway and Zald 2011).Our finding of reduced functional connectivity with the pallidum and putamen further supports their reported association with anhedonia and reduced responsivity to rewards in depression (Pizzagalli et al. 2009; Robinson et al. 2012). Changes involving the right orbitofrontal cortex, putamen, and insula may result in altered responses to negative emotional stimuli in depression and support our previous evidence suggesting altered neural processing of “hate” (insula and putamen) (Zeki and Romaya 2008; Tao et al. 2013).

In schizophrenia, the largest and most extensive functional connectivity changes were found, supporting growing evidence that this disorder is associated with wide-ranging functional dysconnection in the brain (Camchong et al. 2011; Guo et al. 2014). The most extensive circuit affected included the bilateral thalamus, postcentral gyrus, medial frontal gyrus and fusiform gyrus and right dorsal superior frontal gyrus, right putamen and right posterior cingulate gyrus and 18 altered links show a pattern of reciprocal contralateral mediation. Here, the right and left thalamus are particularly prominent suggesting that reduced sensory and motor inputs to and integration with the cortex. Indeed, there is now extensive evidence that thalamic damage and altered connectivity with the cortex in schizophrenia are responsible for cognitive and sensorimotor processing dysfunction (Clinton and Meador-Woodruff 2004; Welsh et al. 2010; Pinault 2011; Marenco et al. 2012). Interestingly, altered links primarily in the medial and frontal regions were associated with positive-symptom severity, with the strength of thalamo-frontal connections being negatively correlated. A similar association was found for functional connections between the right posterior cingulate cortex and left medial frontal cortex. On the other hand thalamic links with motor control regions (postcentral gyrus and rolandic operculum) and the right posterior cortex link to the left fusiform gyrus showed positive correlations with positive-symptom severity. These regions are associated with hallucinations involving different sensory modalities, visual/auditory (left fusiform gyrus), gustatory (rolandic operculum), and somatic (postcentral gyrus) (Weiss and Heckers 1999; Ali et al. 2011).

Negative-symptom severity was negatively correlated with medial cingulate functional connectivity to the pallidum and putamen, which may contribute to altered cognitive, volition, and reward functions in schizophrenia (Foussias and Remington 2010). Posterior cingulate cortex connectivity with visual areas (lingual gyrus and fusiform gyrus) was positively correlated with negative-symptom severity and may reflect compensation for both cognitive and visual processing deficits (Kang et al. 2011). Illness duration correlated with many (6/13) of same functional links associated with negative-symptom severity, but none of those associated with positive ones, perhaps reflecting the progressive worsening of negative symptoms over time (Schultz et al. 1997; Lieberman 1999).

Implications of Altered Contralateral Mediation as a Potential Transdiagnostic Feature

The presence of altered contralateral mediation found in all 3 psychiatric disorders argues for it being a key common motif in a wide range of psychiatric and other mental disorders and therefore an important potential transdiagnostic feature. Overall, SVM analysis showed a good discrimination accuracy of this feature for schizophrenia (82.4%) and slightly less for depression (72.4%) and ADHD (64.7%) although importantly discrimination accuracy was very similar to that using only the Pearson correlation change between functional links (schizophrenia: 84.7%; depression: 85.5%; ADHD: 68%). A recent, large genetic study of mental illnesses suggested that 5 major disorders (schizophrenia, bipolar disorder, autism, major depression, and ADHD) share some common genetic variants (Consortium C-DGotPG 2013) and it is possible that common gene variants might contribute to reduced contralateral hemisphere mediation in mental disorders.

In terms of identifying key molecular targets involved in the control of contralateral mediation, obvious candidate genes would be those both associated with agenesis of the corpus callosum and multiple psychiatric disorders. One potential candidate is the disrupted in schizophrenia gene 1 (DISC1). DISC1 was originally mainly associated with schizophrenia, but intriguingly has recently shown to be associated with callosal agenesis (Osbun et al. 2011) as well as depression, bipolar disorder, and ADHD (Duff et al. 2013; Jacobsen et al. 2013; Thomson et al. 2013). Indeed, in mouse models of schizophrenia where the DISC1 gene is targeted, one of the effects observed is agenesis of the corpus callosum (Jaaro-Peled 2009). Thus, while different patterns of functional-connectivity changes and symptoms occur across different disorders, there may well be common genetic or other factors which lead to altered contralateral mediation by disrupting interhemispheric communication via the corpus callosum, or other fiber tracts such as the anterior commissure. Furthermore, therapeutic strategies involving either transcranial magnetic stimulation, or deep brain stimulation, or voluntary control of brain activity using feedback techniques may benefit from focusing on promoting increased structural and functional connectivity between the hemispheres. Importantly, our findings show that the primary target for such stimulation approaches will often be in the opposite hemisphere to that identified as having altered functional connectivity in patients. This is verified by a recent study showing that endogenous coupling between ipsilesional SMA and M1 was significantly enhanced only after repetitive transcranial magnetic stimulation applied over contralesional M1 in stroke patients (Grefkes et al. 2010), indicating that our findings is potentially useful in treatment and rehabilitation. Indeed, since symmetrical regions often act as mediators for each other, therapeutic strategies using simultaneous or phased bilateral stimulation protocols may be the most efficient.

Funding

J.Z. is funded by National Science Foundation of China NSFC 61004104, 61104143, 61104224 and 11101429. J.F. is a Royal Society Wolfson Research Merit Award holder, partially supported by an EU Grant BION, a UK EPSRC grant and National Centre for Mathematics, the Key Program of National Natural Science Foundation of China (no. 91230201), and Interdisciplinary Sciences (NCMIS) in Chinese Academy of Sciences. K.M.K. was supported by National Natural Science Foundation of China (NSFC) grant 91132720.

Notes

Conflict of Interest: None declared.

References

Ali
S
Patel
M
Avenido
J
Jabeen
S
Riley
WJ
MBA
M
.
2011
.
Hallucinations: common features and causes
.
Curr Psychiatry
 .
10
:
22
.
Anand
A
Li
Y
Wang
Y
Wu
J
Gao
S
Bukhari
L
Mathews
VP
Kalnin
A
Lowe
MJ
.
2005
.
Activity and connectivity of brain mood regulating circuit in depression: a functional magnetic resonance study
.
Biol Psychiatry
 .
57
:
1079
1088
.
Anderson
JS
Druzgal
TJ
Froehlich
A
DuBray
MB
Lange
N
Alexander
AL
Abildskov
T
Nielsen
JA
Cariello
AN
Cooperrider
JR
.
2011
.
Decreased interhemispheric functional connectivity in autism
.
Cereb Cortex
 .
21
:
1134
1146
.
Bearden
CE
van Erp
TG
Dutton
RA
Boyle
C
Madsen
S
Luders
E
Kieseppa
T
Tuulio-Henriksson
A
Huttunen
M
Partonen
T
.
2011
.
Mapping corpus callosum morphology in twin pairs discordant for bipolar disorder
.
Cereb Cortex
 .
21
:
2415
2424
.
Biswal
B
Zerrin Yetkin
F
Haughton
VM
Hyde
JS
.
1995
.
Functional connectivity in the motor cortex of resting human brain using echo-planar mri
.
Magn Reson Med
 .
34
:
537
541
.
Bloom
JS
Hynd
GW
.
2005
.
The role of the corpus callosum in interhemispheric transfer of information: excitation or inhibition?
.
Neuropsychol Rev
 .
15
:
59
71
.
Boes
AD
Bechara
A
Tranel
D
Anderson
SW
Richman
L
Nopoulos
P
.
2009
.
Right ventromedial prefrontal cortex: a neuroanatomical correlate of impulse control in boys
.
Soc Cogn Affect Neurosci
 .
4
:
1
9
.
Buckholtz
JW
Meyer-Lindenberg
A
.
2012
.
Psychopathology and the human connectome: toward a transdiagnostic model of risk for mental illness
.
Neuron
 .
74
:
990
1004
.
Camchong
J
MacDonald
AW
Bell
C
Mueller
BA
Lim
KO
.
2011
.
Altered functional and anatomical connectivity in schizophrenia
.
Schizophr Bull
 .
37
:
640
650
.
Castellanos
FX
Margulies
DS
Kelly
C
Uddin
LQ
Ghaffari
M
Kirsch
A
Shaw
D
Shehzad
Z
Di Martino
A
Biswal
B
.
2008
.
Cingulate-precuneus interactions: a new locus of dysfunction in adult attention-deficit/hyperactivity disorder
.
Biol Psychiatry
 .
63
:
332
337
.
Castellanos
FX
Proal
E
.
2012
.
Large-scale brain systems in ADHD: beyond the prefrontal–striatal model
.
Trends Cogn Sci
 .
16
:
17
26
.
Castellanos
FX
Sonuga-Barke
EJ
Milham
MP
Tannock
R
.
2006
.
Characterizing cognition in ADHD: beyond executive dysfunction
.
Trends Cogn Sci
 .
10
:
117
123
.
Clinton
SM
Meador-Woodruff
JH
.
2004
.
Thalamic dysfunction in schizophrenia: neurochemical, neuropathological, and in vivo imaging abnormalities
.
Schizophr Res
 .
69
:
237
253
.
Compton
RJ
Carp
J
Chaddock
L
Fineman
SL
Quandt
LC
Ratliff
JB
.
2008
.
Trouble crossing the bridge: altered interhemispheric communication of emotional images in anxiety
.
Emotion
 .
8
:
684
.
Consortium C-DGotPG
.
2013
.
Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis
.
Lancet
 .
381
9875
:
1371
1379
.
Crow
T
.
1998
.
Schizophrenia as a transcallosal misconnection syndrome
.
Schizophr Res
 .
30
:
111
114
.
Damoiseaux
JS
Greicius
MD
.
2009
.
Greater than the sum of its parts: a review of studies combining structural connectivity and resting-state functional connectivity
.
Brain Struct Funct
 .
213
:
525
533
.
Di Paola
M
Spalletta
G
Caltagirone
C
.
2010
.
In vivo structural neuroanatomy of corpus callosum in Alzheimer's disease and mild cognitive impairment using different MRI techniques: a review
.
J Alzheimer’s Dis
 .
20
:
67
95
.
Duff
BJ
Macritchie
KA
Moorhead
TW
Lawrie
SM
Blackwood
DH
.
2013
.
Human brain imaging studies of DISC1 in schizophrenia, bipolar disorder and depression: a systematic review
.
Schizophr Res
 .
147
:
1
13
.
Fitzgerald
PB
Laird
AR
Maller
J
Daskalakis
ZJ
.
2008
.
A meta-analytic study of changes in brain activation in depression
.
Hum Brain Mapp
 .
29
:
683
695
.
Foussias
G
Remington
G
.
2010
.
Negative symptoms in schizophrenia: avolition and Occam’s razor
.
Schizophr Bull
 .
36
:
359
369
.
Friston
K
Frith
CD
Fletcher
P
Liddle
P
Frackowiak
R
.
1996
.
Functional topography: multidimensional scaling and functional connectivity in the brain
.
Cereb Cortex
 .
6
:
156
164
.
Frodl
T
Bokde
AL
Scheuerecker
J
Lisiecka
D
Schoepf
V
Hampel
H
Möller
H-J
Brückmann
H
Wiesmann
M
Meisenzahl
E
.
2010
.
Functional connectivity bias of the orbitofrontal cortex in drug-free patients with major depression
.
Biol Psychiatry
 .
67
:
161
167
.
Garrity
A
Pearlson
G
McKiernan
K
Lloyd
D
Kiehl
K
Calhoun
V
.
2007
.
Aberrant “default mode” functional connectivity in schizophrenia
.
Am J Psychiatry
 .
164
:
450
457
.
Gilliam
M
Stockman
M
Malek
M
Sharp
W
Greenstein
D
Lalonde
F
Clasen
L
Giedd
J
Rapoport
J
Shaw
P
.
2011
.
Developmental trajectories of the corpus callosum in attention-deficit/hyperactivity disorder
.
Biol Psychiatry
 .
69
:
839
846
.
Grefkes
C
Nowak
DA
Wang
LE
Dafotakis
M
Eickhoff
SB
Fink
GR
.
2010
.
Modulating cortical connectivity in stroke patients by rTMS assessed with fMRI and dynamic causal modeling
.
Neuroimage
 .
50
:
233
242
.
Greicius
MD
Flores
BH
Menon
V
Glover
GH
Solvason
HB
Kenna
H
Reiss
AL
Schatzberg
AF
.
2007
.
Resting-state functional connectivity in major depression: abnormally increased contributions from subgenual cingulate cortex and thalamus
.
Biol Psychiatry
 .
62
:
429
437
.
Guo
S
Kendrick
KM
Yu
R
Wang
HLS
Feng
J
.
2014
.
Key functional circuitry altered in schizophrenia involves parietal regions associated with sense of self
.
Hum Brain Mapp
 .
35
:
123
139
.
Guo
S
Kendrick
KM
Zhang
J
Broome
M
Yu
R
Liu
Z
Feng
J
.
2013
.
Brain-wide functional inter-hemispheric disconnection is a potential biomarker for schizophrenia and distinguishes it from depression
.
NeuroImage Clin
 .
2
:
818
826
.
Hesslinger
B
Tebartz van Elst
L
Thiel
T
Haegele
K
Hennig
J
Ebert
D
.
2002
.
Frontoorbital volume reductions in adult patients with attention deficit hyperactivity disorder
.
Neurosci Lett
 .
328
:
319
321
.
Honey
C
Sporns
O
Cammoun
L
Gigandet
X
Thiran
J-P
Meuli
R
Hagmann
P
.
2009
.
Predicting human resting-state functional connectivity from structural connectivity
.
Proc Natl Acad Sci
 .
106
:
2035
2040
.
Jaaro-Peled
H
.
2009
.
Gene models of schizophrenia: DISC1 mouse models
.
Prog Brain Res
 .
179
:
75
86
.
Jacobsen
KK
Halmøy
A
Sánchez-Mora
C
Ramos-Quiroga
JA
Cormand
B
Haavik
J
Johansson
S
.
2013
.
DISC1 in adult ADHD patients: An association study in two European samples
.
Am J Med Genet B Neuropsychiatr Genet
 .
162
:
227
234
.
Kang
SS
Sponheim
SR
Chafee
MV
MacDonald III
AW
.
2011
.
Disrupted functional connectivity for controlled visual processing as a basis for impaired spatial working memory in schizophrenia
.
Neuropsychologia
 .
49
:
2836
2847
.
Knöchel
C
Oertel-Knöchel
V
Schönmeyer
R
Rotarska-Jagiela
A
van de Ven
V
Prvulovic
D
Haenschel
C
Uhlhaas
P
Pantel
J
Hampel
H
.
2012
.
Interhemispheric hypoconnectivity in schizophrenia: fiber integrity and volume differences of the corpus callosum in patients and unaffected relatives
.
Neuroimage
 .
59
:
926
934
.
Liang
M
Zhou
Y
Jiang
T
Liu
Z
Tian
L
Liu
H
Hao
Y
.
2006
.
Widespread functional disconnectivity in schizophrenia with resting-state functional magnetic resonance imaging
.
Neuroreport
 .
17
:
209
213
.
Lieberman
JA
.
1999
.
Is schizophrenia a neurodegenerative disorder? A clinical and neurobiological perspective
.
Biol Psychiatry
 .
46
:
729
739
.
Liu
Y
Liang
M
Zhou
Y
He
Y
Hao
Y
Song
M
Yu
C
Liu
H
Liu
Z
Jiang
T
.
2008
.
Disrupted small-world networks in schizophrenia
.
Brain
 .
131
:
945
961
.
Lui
S
Wu
Q
Qiu
L
Yang
X
Kuang
W
Chan
RC
Huang
X
Kemp
GJ
Mechelli
A
Gong
Q
.
2011
.
Resting-state functional connectivity in treatment-resistant depression
.
Am J Psychiatry
 .
168
:
642
648
.
Lynall
M-E
Bassett
DS
Kerwin
R
McKenna
PJ
Kitzbichler
M
Muller
U
Bullmore
E
.
2010
.
Functional connectivity and brain networks in schizophrenia
.
J Neurosci
 .
30
:
9477
9487
.
Marenco
S
Stein
JL
Savostyanova
AA
Sambataro
F
Tan
H-Y
Goldman
AL
Verchinski
BA
Barnett
AS
Dickinson
D
Apud
JA
.
2012
.
Investigation of anatomical thalamo-cortical connectivity and FMRI activation in schizophrenia
.
Neuropsychopharmacology
 .
37
:
499
507
.
Marrelec
G
Kim
J
Doyon
J
Horwitz
B
.
2009
.
Large scale neural model validation of partial correlation analysis for effective connectivity investigation in functional MRI
.
Hum Brain Mapp
 .
30
:
941
950
.
Marrelec
G
Krainik
A
Duffau
H
Pélégrini-Issac
M
Lehéricy
S
Doyon
J
Benali
H
.
2006
.
Partial correlation for functional brain interactivity investigation in functional MRI
.
Neuroimage
 .
32
:
228
237
.
Mazaheri
A
Coffey-Corina
S
Mangun
GR
Bekker
EM
Berry
AS
Corbett
BA
.
2010
.
Functional disconnection of frontal cortex and visual cortex in attention-deficit/hyperactivity disorder
.
Biol Psychiatry
 .
67
:
617
623
.
Mills
KL
Bathula
D
Dias
TGC
Iyer
SP
Fenesy
MC
Musser
ED
Stevens
CA
Thurlow
BL
Carpenter
SD
Nagel
BJ
.
2012
.
Altered cortico-striatal–thalamic connectivity in relation to spatial working memory capacity in children with ADHD
.
Front Psychiatry
 .
3
2
:
1
17
.
Müller
R-A
Shih
P
Keehn
B
Deyoe
JR
Leyden
KM
Shukla
DK
.
2011
.
Underconnected, but how? A survey of functional connectivity MRI studies in autism spectrum disorders
.
Cereb Cortex
 .
21
:
2233
2243
.
Osbun
N
Li
J
O'Driscoll
MC
Strominger
Z
Wakahiro
M
Rider
E
Bukshpun
P
Boland
E
Spurrell
CH
Schackwitz
W
.
2011
.
Genetic and functional analyses identify DISC1 as a novel callosal agenesis candidate gene
.
Am J Med Genet A
 .
155
:
1865
1876
.
Ovadia-Caro
S
Nir
Y
Soddu
A
Ramot
M
Hesselmann
G
Vanhaudenhuyse
A
Dinstein
I
Tshibanda
J-FL
Boly
M
Harel
M
.
2012
.
Reduction in inter-hemispheric connectivity in disorders of consciousness
.
PLoS ONE
 .
7
:
e37238
.
Pinault
D
.
2011
.
Dysfunctional thalamus-related networks in schizophrenia
.
Schizophr Bull
 .
37
:
238
243
.
Pizzagalli
D
Holmes
A
Dillon
D
Goetz
E
Birk
J
Bogdan
R
Dougherty
D
Iosifescu
D
Rauch
S
Fava
M
.
2009
.
Reduced caudate and nucleus accumbens response to rewards in unmedicated individuals with major depressive disorder
.
Am J Psychiatry
 .
166
:
702
710
.
Plessen
KJ
Wentzel-Larsen
T
Hugdahl
K
Feineigle
P
Klein
J
Staib
LH
Leckman
JF
Bansal
R
Peterson
BS
.
2004
.
Altered interhemispheric connectivity in individuals with Tourette's disorder
.
Am J Psychiatry
 .
161
:
2028
2037
.
Robbins
TW
Gillan
CM
Smith
DG
de Wit
S
Ersche
KD
.
2012
.
Neurocognitive endophenotypes of impulsivity and compulsivity: towards dimensional psychiatry
.
Trends Cogn Sci
 .
16
:
81
91
.
Robinson
OJ
Cools
R
Carlisi
CO
Sahakian
BJ
Drevets
WC
.
2012
.
Ventral striatum response during reward and punishment reversal learning in unmedicated major depressive disorder
.
Am J Psychiatry
 .
169
:
152
159
.
Rolls
ET
.
2004
.
The functions of the orbitofrontal cortex
.
Brain Cogn
 .
55
:
11
29
.
Rüsch
N
Bracht
T
Kreher
BW
Schnell
S
Glauche
V
Il'yasov
KA
Ebert
D
Lieb
K
Hennig
J
Saur
D
.
2010
.
Reduced interhemispheric structural connectivity between anterior cingulate cortices in borderline personality disorder
.
Psychiatry Res Neuroimaging
 .
181
:
151
154
.
Scheuerecker
J
Meisenzahl
EM
Koutsouleris
N
Roesner
M
Schöpf
V
Linn
J
Wiesmann
M
Brückmann
H
Möller
H-J
Frodl
T
.
2010
.
Orbitofrontal volume reductions during emotion recognition in patients with major depression
.
J Psychiatry Neurosci
 .
35
:
311
.
Schultz
SK
Miller
DD
Oliver
SE
Arndt
S
Flaum
M
Andreasen
NC
.
1997
.
The life course of schizophrenia: age and symptom dimensions
.
Schizophr Res
 .
23
:
15
23
.
Sheline
YI
Raichle
ME
.
2013
.
Resting state functional connectivity in preclinical Alzheimer's disease
.
Biol Psychiatry
 .
74
:
340
347
.
Sylvester
C
Corbetta
M
Raichle
M
Rodebaugh
T
Schlaggar
B
Sheline
Y
Zorumski
C
Lenze
E
.
2012
.
Functional network dysfunction in anxiety and anxiety disorders
.
Trends Neurosci
 .
35
:
527
535
.
Tao
H
Guo
S
Ge
T
Kendrick
KM
Xue
Z
Liu
Z
Feng
J
.
2013
.
Depression uncouples brain hate circuit
.
Mol Psychiatry
 .
18
:
101
111
.
Thomson
P
Macintyre
D
Hamilton
G
Dominiczak
A
Smith
B
Morris
A
Evans
K
Porteous
D
.
2013
.
Association of DISC1 variants with age of onset in a population-based sample of recurrent major depression
.
Mol Psychiatry
 .
18
:
745
747
.
Treadway
MT
Zald
DH
.
2011
.
Reconsidering anhedonia in depression: lessons from translational neuroscience
.
Neurosci Biobehav Rev
 .
35
:
537
555
.
Tzourio-Mazoyer
N
Landeau
B
Papathanassiou
D
Crivello
F
Etard
O
Delcroix
N
Mazoyer
B
Joliot
M
.
2002
.
Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain
.
Neuroimage
 .
15
:
273
289
.
Uddin
LQ
Kelly
A
Biswal
BB
Margulies
DS
Shehzad
Z
Shaw
D
Ghaffari
M
Rotrosen
J
Adler
LA
Castellanos
FX
.
2008
.
Network homogeneity reveals decreased integrity of default-mode network in ADHD
.
J Neurosci Methods
 .
169
:
249
254
.
van der Knaap
LJ
van der Ham
IJ
.
2011
.
How does the corpus callosum mediate interhemispheric transfer?
Rev Behav Brain Res
 .
223
:
211
221
.
Wang
K
Liang
M
Wang
L
Tian
L
Zhang
X
Li
K
Jiang
T
.
2007
.
Altered functional connectivity in early Alzheimer’s disease: a resting-state fMRI study
.
Hum Brain Mapp
 .
28
:
967
978
.
Weiss
AP
Heckers
S
.
1999
.
Neuroimaging of hallucinations: a review of the literature
.
Psychiatry Res Neuroimaging
 .
92
:
61
74
.
Welsh
RC
Chen
AC
Taylor
SF
.
2010
.
Low-frequency BOLD fluctuations demonstrate altered thalamocortical connectivity in schizophrenia
.
Schizophr Bull
 .
36
:
713
722
.
Whitfield-Gabrieli
S
Thermenos
HW
Milanovic
S
Tsuang
MT
Faraone
SV
McCarley
RW
Shenton
ME
Green
AI
Nieto-Castanon
A
LaViolette
P
.
2009
.
Hyperactivity and hyperconnectivity of the default network in schizophrenia and in first-degree relatives of persons with schizophrenia
.
Proc Natl Acad Sci
 .
106
:
1279
1284
.
Xu
K
Jiang
W
Ren
L
Ouyang
X
Jiang
Y
Wu
F
Kong
L
Womer
F
Liu
Z
Blumberg
HP
.
2013
.
Impaired interhemispheric connectivity in medication-naive patients with major depressive disorder
.
J Psychiatry Neurosci
 .
38
:
43
.
Zeki
S
Romaya
JP
.
2008
.
Neural correlates of hate
.
PLoS ONE
 .
3
:
e3556
.
Zeng
L-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
Snyder
AZ
Shimony
JS
Fox
MD
Raichle
ME
.
2010
.
Noninvasive functional and structural connectivity mapping of the human thalamocortical system
.
Cereb Cortex
 .
20
:
1187
1194
.

Author notes

Jie Zhang and Keith M. Kendrick contributed equally to the work and are joint first authors.