Modulation of functional network co-activation pattern dynamics following ketamine treatment in major depression

Abstract

Ketamine produces fast-acting antidepressant effects in treatment-resistant depression (TRD). Prior studies have shown altered functional dynamics between brain networks in major depression. We thus sought to determine whether functional brain network dynamics are modulated by ketamine therapy in TRD. Participants with TRD (n = 58, mean age = 40.7 years, female = 48.3%) completed resting-state fMRI scans and clinical assessments (mood and rumination) at baseline and 24 h after receiving 4 ketamine infusions (0.5 mg/kg) over 2 weeks. Healthy controls (HC) (n = 56, mean age = 32.8 years, female = 57.1%) received the same assessments at baseline and after 2 weeks in a subsample without treatment. A co-activation pattern (CAP) analysis identified recurring patterns of brain activity across all subjects using k-means clustering. Statistical analyses compared CAP metrics including the fraction of time (FT) spent in a brain state, and the transition probability (TP) from one state to another over time and associations with clinical improvement. Follow-up analyses compared HC and TRD at baseline. Six brain state clusters were identified, including patterns resembling the salience (SN), central executive (CEN), visual (VN), default mode (DMN), and somatomotor (SMN) networks. Following ketamine treatment, TRD patients showed decreased FT for the VN (p = 7.4E-04) and increased FT for the CEN state (p = 1.9E-03). For TP metrics, SN-CEN increased (p = 5.8E-04) and SN-VN decreased (p = 3.6E-03). Decreased FT for the SN associated with improved rumination (p = 1.9E-03). At baseline, lower FT for CEN (p = 5.70E-04) and TP for SN-CEN (p = 0.016) and higher TP for SN-VN (p = 2.60E-03) distinguished TRD from HCs. CAP metrics remained stable over time in a subsample of HCs (n = 18). These findings suggest ketamine modulates brain network dynamics between SN, CEN, and VN in TRD, which may normalize dynamic patterns seen in TRD at baseline toward patterns seen in controls. Changes in SN state dynamics may correspond to improvements in ruminative symptoms following ketamine therapy.

Keywords: central executive network; co-activation patterns; dynamic functional connectivity; ketamine; rumination; salience network; treatment-resistant depression.

Fig. 1.

Overview of study design. At each time point, all participants received the same set of MRI scans and clinical assessments. Treatment-resistant depression participants (TRD) (n = 58) were scanned at baseline and after receiving 4 serial ketamine infusions over a period of about 2 weeks. Healthy controls (HC) were scanned at baseline, with a small subset (n = 18) being scanned again after 2 weeks, following a similar time frame to the period of time ketamine was administered to TRD.

Fig. 2.

CAP states revealed from clustering fMRI data across all subjects and scan sessions. Custer analyses revealed six states as optimal in representing dynamic brain states. (Top) The centroid of each cluster was projected back onto the brain in CIFTI space to visualize the spatial activation patterns of each brain state, shown on an inflated cortical surface and subcortical volume structures below. Colors are represented as z-scores to show the relative strength of activation or deactivation across brain network nodes. State 1 is characterized by activation in the visual (VN) and dorsal attention (DAN) networks, with moderate activation in the amygdala. State 2 is characterized by activation in the somatomotor network (SMN), with moderate amygdala and hippocampus activation. State 3 is characterized by activation in the canonical default mode network (DMN), amygdala, and hippocampus. State 4 is characterized by activation in ventral and lateral regions of the DMN that are not as strongly represented in state 3, with moderate VN, amygdala, and striatum activation. State 5 is characterized by activation in the salience network (SN), with moderate striatum activation. State 6 is characterized by activation in the central executive network (CEN), weak activation in some medial DMN regions, and moderate activation in the striatum. (Bottom) Each CAP state (rows) was correlated with the set of parcels belonging to each of the seven network states from the Schaefer atlas (columns) in order to quantitatively determine the networks in each CAP state. The correlations show correspondence with the network descriptions based on the visualizations of each CAP state.

Fig. 3.

Changes in CAP metrics over time following SKI. Each set of boxplots shows a distribution of values for CAP metrics corresponding to brain states in the TRD sample at baseline (TRD_Basline), following ketamine treatment (TRD_Post_SKI), and in controls (HC). Stars above the boxplots indicate the level of significance for each test performed across the groups (*: p < 0.05, **: p < 0.01, ***: p < 0.001). To the left of each boxplot is a visual representation of the brain states implicated, which include a single brain state or a transition between brain states. Following ketamine treatment, we observed significant decreases in the fraction of time spent in a visual network (VN) state (top left), significant increases in the fraction of time spent in a central executive network (CEN) state (bottom left), significant decreases in the transition probability between a salience network (SN) state and VN state (top right), and significant increases in the transition probability between the SN state and CEN state (bottom right). Follow-up analyses that tested for significant differences between controls and TRD at baseline revealed significantly lower fraction of time spent in the CEN, significantly higher transition probability between SN and VN, and significantly lower transition probability between SN and CEN in TRD compared with controls.

Fig. 4.

Associations with changes in rumination following ketamine treatment. Percentage improvement in reflective rumination from the RRS measure was significantly correlated with decreases in the fraction of time spent in the salience network (SN) state. As a follow-up, cross-sectional analysis was performed to determine whether trends toward improving symptoms following ketamine treatment moved toward a pattern seen in controls to show TRD participants spent a higher fraction of time in the SN state compared with controls. Scatter plot showing the relationship between SN state dynamics and RRS is displayed on the right, boxplots showing differences between TRD and controls are displayed in the middle, and a visualization of the SN state is displayed on the right. Significant cross-sectional differences imply that FT of CAP 5 in subjects which show greater RRS improvements trend toward FT values seen in HC. *p < 0.05, **p < 0.01, ***p < 0.001.