Although major depression has been considered as a manifestation of discoordinated activity between affective and cognitive neural networks, only a few studies have examined the relationships among neural networks directly. Because of the known disconnection theory, geriatric depression could be a useful model in studying the interactions among different networks. In the present study, using independent component analysis to identify intrinsically connected neural networks, we investigated the alterations in synchronizations among neural networks in geriatric depression to better understand the underlying neural mechanisms. Resting-state fMRI data was collected from thirty-two patients with geriatric depression and thirty-two age-matched never-depressed controls. We compared the resting-state activities between the two groups in the default-mode, central executive, attention, salience, and affective networks as well as correlations among these networks. The depression group showed stronger activity than the controls in an affective network, specifically within the orbitofrontal region. However, unlike the never-depressed controls, geriatric depression group lacked synchronized/antisynchronized activity between the affective network and the other networks. Those depressed patients with lower executive function has greater synchronization between the salience network with the executive and affective networks. Our results demonstrate the effectiveness of the between-network analyses in examining neural models for geriatric depression.
It has long been postulated that major depression may be a consequence of failed coordination between the central executive system and affective processing system [
In addition to the executive and affective processing systems, the abnormalities in the default-mode network (DMN, primarily including the anterior and posterior cingulate, and bilateral lateral parietal cortex areas) and salience network (including the dorsal anterior cingulate and insula cortices) in major depression have also been identified [
Major depression in individuals who had the first depression episode at their older ages (typically older than 50 years) is often referred as geriatric depression. Different from major depression in younger adults, geriatric depression has frequently been found in those with cerebrovascular disorders [
In recent years, with the development of various analyzing methods on task-related and task-free functional magnetic resonance imaging (fMRI) data, analyzing fMRI data at a neural network level becomes a reality. One of the widely used techniques to identify neural networks is the independent component analysis (ICA). Unlike the seed-based functional connectivity analysis which is dependent on the location and size of a seed, the ICA approach is data driven. It identifies independent components (ICs) based on the spatial and temporal distribution patterns [
The majority of previous studies in the literature have examined the association of regional activity or connectivity between two regions with depression severity [
Thirty-two individuals who had been diagnosed with major depressive disorder (19 females, mean ± SD age:
The clinical profiles of the participants.
Depression ( |
Control ( |
|
|
---|---|---|---|
Gender (F/M) | 18/14 | 19/13 | 0.80+ |
Age | 68.3 (6.5) | 71.8 (8.2) | 0.06 |
Years of education | 14.9 (3.1) | 16.0 (2.5) | 0.11 |
MADRS | 7.0 (9.2) | 0.0 (0.9) | <0.001* |
Number medicated for hypotension | 11 | 8 | 0.40+ |
Number medicated for antidepressants | 18 | 0 | |
Monotherapy | |||
SSRI | 4 | ||
SARI | 2 | ||
SNRI | 1 | ||
Tricyclic | 2 | ||
Combined treatment | |||
Two SSRIs | 4 | ||
SSRI with either SARI or NDRI | 2 | ||
SARI & NDRI | 2 | ||
SNRI & NDRI | 1 | ||
Executive function (Stroop task) | −0.10 (0.82) | 0.24 (0.71) | 0.08 |
Prior to the fMRI, all subjects completed the Stroop Color and Word Test to examine the executive function. The study received approval by Duke School of Medicine Institutional Review Board. All subjects gave verbal and written consent after being explained the purpose and procedures to be used in the study.
All participants were scanned using a research-dedicated 3.0 T GE EXCITE HD scanner (GE Medical Systems, Milwaukee, Wisconsin). First, high-resolution T1-weighted structural images in coronel view were acquired with slice thickness of 1 mm without a gap (matrix = 256 × 256 × 216). We then obtained 5-minute resting fMRI scans for each participant. Participants were instructed to rest without moving, keep their eyes open, and focus on a fixation cross-presented in the center of the screen inside the scanner. Inward spiral sequence functional images in the axial view were acquired using the following parameters: TR = 2000 ms, TE = 31 ms, FOV = 24 cm, flip angle = 90°, and matrix = 64 × 64 × 34.
Data were preprocessed using the Duke BIAC resting state pipeline based on the tools from the FSL analysis package (FMRIB Software Library,
To identify group difference between the depression patients and controls in each component identified as DMN, CEN, CAN, SN, and AN and the inter-IC correlations between any two networks, voxelwise two-sample
We summarize the demographic details, clinical profile, and performance of participants on the Stroop task in Table
Our aim was to examine whether we can identify altered interactions among networks that are related to depressive symptoms and cognitive dysfunctions in geriatric depression. To achieve this goal, using the results from Laird and colleagues [
The clusters of each IC component identified matches the CEN, CAN, DMN, AN, and SN, respectively.
Network | IC | Clusters | Peak coordinate (MNI |
---|---|---|---|
DMN | IC1 | Bilateral medial prefrontal cortex |
|
Bilateral posterior cingulate |
|
||
Bilateral lateral parietal cortex |
|
||
|
|||
AN | IC12 | Bilateral dorsomedial prefrontal cortex |
|
Rostral anterior cingulate |
|
||
Bilateral subgenual cingulate |
|
||
IC18 | Bilateral subgenual cingulate |
|
|
Bilateral rostral anterior cingulate |
|
||
Bilateral ventrolateral prefrontal cortex |
|
||
Bilateral orbitofrontal cortex |
|
||
Bilateral amygdala |
|
||
Bilateral caudate |
|
||
|
|||
CEN | IC6 | Left dorsolateral prefrontal cortex |
|
Left dorsomedial prefrontal cortex |
|
||
Bilateral superior parietal cortex |
|
||
Right inferior temporal cortex |
|
||
Left anterior part of posterior cingulate |
|
||
Right cerebellum |
|
||
IC4 | Right dorsolateral prefrontal cortex |
|
|
Right dorsomedial prefrontal cortex |
|
||
Bilateral superior parietal cortex |
|
||
Right inferior temporal cortex |
|
||
Right anterior part of posterior cingulate |
|
||
Left cerebellum |
|
||
|
|||
CAN | IC7 | Bilateral frontal eye field |
|
Bilateral precuneus |
|
||
Bilateral parieto-occipital fissure |
|
||
Bilateral lingual gyrus |
|
||
|
|||
SN | IC10 | Bilateral dorsal cingulate |
|
Bilateral insula |
|
||
Bilateral parieto-occipital fissure |
|
The ICA components which correspond to different neural networks according to the goodness-of-fit analysis using the templates of Laird et al. AN = affective network; CAN = central attentional network; CEN = central executive network; DMN = default-mode network; SN = salience network.
The locations of each IC component which correspond to different neural networks according to the goodness-of-fit analysis using the templates of Laird et al. The IC components were computed using dual regression analysis by combining the data from both the depression and never-depressed control group (
When comparing each individual network between patients and controls using two-sample
(a) Component 12 (IC12), one of the affective networks, in the control group; (b) IC12 in the depression group. To show the voxels in the cerebellum, (a) and (b) were based on threshold of
When comparing correlations among networks between the two groups (pairwise correlations), it was the IC12 that showed a significant group difference in the synchronizations between this network with several other networks (Table
Mean (SD) inter-IC correlations that showed significant group differences and that correlated with performance of the Stroop task.
Control | Depression |
|
|
---|---|---|---|
IC12(AN)-IC6(CEN) correlation | 0.30 (0.28) | 0.15 (0.26) | 0.03 |
IC12(AN)-IC7(CAN) correlation | −0.39 (0.33) | −0.22 (0.32) | 0.04 |
IC12(AN)-IC10(SN) correlation | −0.18 (0.29) | −0.02 (0.32) | 0.04 |
Stroop performance with IC10(SN)-IC6(CEN) correlation |
|
0.05 | |
Stroop performance with IC10(SN)-IC18(AN) correlation |
|
0.003 |
(a) Neural network pairs which revealed significant correlations in the control group. (b) The plots of time course of each network in a control subject (ID10) and a depression subject (ID37) to illustrate the interaction effect between each paired neural networks. The significance was tested using Monte-Carlo simulation.
We did not find any significant correlation of the severity of depressive symptoms with any of the networks or interactions between any two networks. However, as shown in Figure
(a) Neural network pairs that their correlation or anticorrelation was correlated with the Stroop task performance in depression patients. (b) The correlation plots in the patient group (red) and control group (blue).
Given the fact that we had a fair number of patients in a remitted state, we suspected that the reason that we did not find a significant correlation between neural activity and depression severity might be because their relationship is nonlinear; that is, it might be a depression-state dependent rather than a linear relationship. Therefore, we subsequently examined the differences in neural networks and interactions of networks between the remitted versus the actively depressed groups and between the remitted versus the control groups. As shown in Figures
Upper: the IC18 in the control group (a) and depression group. (b) Lower: the regions within IC18 that revealed significantly increased activity in the depressed patients than controls (c) and in the actively depressed patients than the remitted patients (d) (
We also examined the group differences in network synchronizations between the actively depressed versus control, actively depressed versus remitted, and remitted versus control groups. The analyses confirmed the network interaction results in the combined patient sample, in that the significant positive correlation between IC10 (SN) and IC6 (CEN) and the negative correlation between IC10 (SN) and IC7 (CAN) in the control group were significantly less positive or less negative in the actively depressed group. There were no significant group differences between the actively depressed and remitted groups or between the remitted and control groups among the network interactions.
We investigated the interactions among different intrinsic connectivity networks in patients with both acute and remitted geriatric depression and found that depression patients had significant alterations in the synchronizations/antisynchronizations between the affective network with other networks including the central executive network, attentional network, and the salience network. In addition, we found depressive-state specific increase in the orbitofrontal area of the affective network. Although these changes were not correlated with depression severity, the significant differences confirmed in the acutely depressed group indicate an importance of the interactions between networks as the neuropathology of major depression.
It is interesting that the depression group mainly had altered correlations between the component of the affective network (including the orbitofrontal, subgenual cingulate, and the dorsomedial prefrontal cortex) with other neural networks. This component best matched component 2 (subgenual cingulate and orbitofrontal cortex) of Laird and colleagues’ study and the authors indicated its role in “olfaction, gustation, and emotion” [
We also found a negative correlation between the IC12 with IC7 and IC10 in the healthy control group and the correlation become less negative in the depressed group. The IC7 located at the frontal eye area and the precuneus areas and best matched the IC7 of Laird’s study which should be an attentional network, whereas IC10 matched the IC4 of Laird’s study (the bilateral anterior insula/frontal opercula and ACC) which should be the salience network. The salience network recently has been hypothesized to play an important role in facilitating attentional transition between cognition and emotion/interoception [
Since we mainly found a discoordination between the affective networks with other networks, we speculate that the primary deficits in depression could be in the automatic emotion regulation system of the affective network which have resulted in the interaction deficits between this network and other networks. Indeed, we found increased activity in the left orbitofrontal cortex area (although not IC12 but IC18 instead) in depression, especially in the actively depressed group relative to both the remitted depression group and healthy control group (suggesting a depressive state-related alteration). As shown in the results, there are some spatial overlaps in the ventromedial prefrontal and orbitofrontal regions between the IC18 and IC12. Similar to IC12, the IC18 also matched the IC2 as well as IC1 of Lairds’ study, but the IC18 also included the limbic and brainstem regions, all of which should also be part of the automatic emotion regulation network. The pathological deficit in the orbitofrontal cortex in depression has long been well documented [
In this study, we also found a negative correlation between the Stroop task performance and the synchronization of the salience network (IC10) with the central executive network/voluntary emotion regulation network (IC6) as well as the synchronization of the salience network with the automatic emotion regulation network (IC18) in the depression group but not in the control group. In other words, those patients, who had poor performance in the Stroop task, had stronger synchronization between the salience network and the emotional regulation (both voluntary and automatic) networks. The results may implicate that, those whose salience network and emotional regulation networks are positively synchronized, may be more easily to reallocate their attention to emotional events, which then could distract them from ongoing cognitive tasks and result in poor performance in the executive tasks such as the Stroop task. Supportively, in the control group, the salience network was negatively correlated with the automatic emotion regulation network (IC12 though). However, it is difficult to explain why those who had poor performance during the Stroop task had a positive correlation between the salience network and the voluntary emotional regulation network. While it might be a compensatory effect, future studies to investigate the synchronization between the salience network and the voluntary emotional regulation (central executive) network during performing the Stroop task scan is necessary to explain the phenomenon.
There are a couple of technical issues that need to be discussed here. First of all, we conducted motion correction during preprocessing which has not been frequently reported in the literature in ICA analyses. We rationale that for the majority of functional connectivity analysis (e.g., seed-based connectivity analysis), in addition to the slice-time correction, motion correction, and normalization, filtering and regressing out covariates (such as six motion parameters, white matter signal, and CSF signal) are also essential during data preprocessing [
In addition, we only studied the interactions between any two neural networks at a time. Using more complicated models that calculate the interactions among the networks simultaneously is necessary to confirm our results. We studied the internetwork correlations in geriatric depression because there have been known pathological disconnections in geriatric depression. Thus, our results cannot be generalized to younger depression patients. Because different regions might be involved in the pathology of geriatric depression due to large variations of outcomes from cerebrovascular diseases, further studies should be conducted to examine whether and how different cerebrovascular deficits affect our findings. This study is also limited by the small sample size and different medications of the depression patients. The small number of actively depressed patients may impact on the robustness of the significance of our results. This might explain why we only found significant alterations in the affective network but not in the executive network. Based on the results from the small number of patients, perhaps what we may conclude here is that at least deficits in the affective network were more robust and obvious than those in the executive network in the actively depressed group. Future replication studies in unmedicated patients with geriatric depression in a larger sample are warranted to confirm our conclusions.
While deficits of resting activity in depression have been reported in a number of studies in major depression, the aberrant interactions among intrinsic neural networks have not been demonstrated previously. Although our current study cannot determine which was the primary deficit in major depression firmly, the altered network activity, or the interactions among networks, we were able to examine the interactions between networks directly using the ICA approach. Our results have demonstrated that hyperactivity within the affective network (the automatic emotion regulation system), in particular the orbitofrontal cortex, in conjunction with sparse correlation among the central executive network, attentional network, and the salience network, is the core dysfunction of older depression patients during resting state. The results are in consistent with several depression models proposed in the literature and indicated that studying the correlations among networks is an effective approach in revealing neural mechanisms of depression.
The authors declare that there is no conflict of interests regarding the publication of this paper.
This research was supported by the Paul B. Beeson Career Developmental Awards (K23-AG028982). Lihong Wang and Ying-Hui Chou are supported by NIMH R01MH098301-01A1. David C. Steffens is supported by an NIMH R01 MH054846. Guy G. Potter is supported by NIMH Career Development Award (K23 MH087741).