Introduction
Repetitive negative thinking (RNT), such as rumination in the context of depression, is a cognitive process characterized by a persistent focus on negative experiences related to the self (Nolen-Hoeksema, Wisco, & Lyubomirsky, Reference Nolen-Hoeksema, Wisco and Lyubomirsky2008). RNT is a symptom dimension with significant implications for the course and prognosis of depression, making this disorder refractory to treatment, chronic, and complicated with suicide (Krajniak, Miranda, & Wheeler, Reference Krajniak, Miranda and Wheeler2013; Surrence, Miranda, Marroquin, & Chan, Reference Surrence, Miranda, Marroquin and Chan2009; Watkins & Roberts, Reference Watkins and Roberts2020). Previous research has examined the triggers, intensity, and duration of RNT. Characterizing the neurobiological mechanisms of RNT is important not only for understanding its formation but also for identifying neuromodulation targets aimed at alleviating this symptom.
Prior functional connectivity (FC)-based studies have identified many regions of interest (ROIs) as they relate to heightened RNT and brooding symptoms in individuals, including the left dorsolateral prefrontal cortex, precuneus, and other components of the default mode network (DMN) (Jacob et al., Reference Jacob, Morris, Huang, Schneider, Rutter, Verma, Murrough and Balchandani2020; Taylor et al., Reference Taylor, Yamada, Kawashima, Kobayashi, Yoshihara, Miyata, Murai, Kawato and Motegi2022). However, our previous resting-state fMRI study revealed that RNT intensity correlates with increased FC between the bilateral anterior insular cortices and the right superior temporal sulcus (STS) (Tsuchiyagaito et al., Reference Tsuchiyagaito, Sanchez, Misaki, Kuplicki, Park, Paulus and Guinjoan2022). This result highlighted the neural mechanisms underlying RNT as difficulties in disengaging attention from negative emotional responses (Craig, Reference Craig2009), and having interrelation with inner-speech processing (Deen, Koldewyn, Kanwisher, & Saxe, Reference Deen, Koldewyn, Kanwisher and Saxe2015). This is compatible with the view that the DMN serves resting self-dialogue, but not necessarily depressive rumination (Goldstein-Piekarski et al., Reference Goldstein-Piekarski, Ball, Samara, Staveland, Keller, Fleming, Grisanzio, Holt-Gosselin, Stetz and Ma2022). Thus, prior evidence deemphasizes the role of DMN dysfunction in RNT (Goldstein-Piekarski et al., Reference Goldstein-Piekarski, Ball, Samara, Staveland, Keller, Fleming, Grisanzio, Holt-Gosselin, Stetz and Ma2022; Makovac et al., Reference Makovac, Fagioli, Rae, Critchley and Ottaviani2020; Tozzi et al., Reference Tozzi, Zhang, Chesnut, Holt-Gosselin, Ramirez and Williams2021), while recent work by our group (Tsuchiyagaito et al., Reference Tsuchiyagaito, Sanchez, Misaki, Kuplicki, Park, Paulus and Guinjoan2022) demonstrates that the functional connection between the insula (Craig, Reference Craig2009) and the STS (Deen et al., Reference Deen, Koldewyn, Kanwisher and Saxe2015) is related to the intensity of RNT (Tsuchiyagaito et al., Reference Tsuchiyagaito, Sanchez, Misaki, Kuplicki, Park, Paulus and Guinjoan2022). Nevertheless, our understanding is limited to the resting-state data, which lack clarity on the RNT circuit when individuals are actively engaging with RNT.
RNT has been established as a trait-like cognitive process that involves a recurrent and continuous focus on self-relevant negative thoughts that are persistent over time and across situations. However, RNT intensity can also fluctuate, such that there is a state component to it; it can be influenced by overall depression symptom severity, instant mood state, and adverse environmental stimuli – including relevant interpersonal interactions (Chang et al., Reference Chang, Berenz, Ajilore, Langenecker, Burgess, Phan and Klumpp2023; Philippi et al., Reference Philippi, Leutzinger, Pessin, Cassani, Mikel, Walsh, Hoks, Birn and Abercrombie2022). This differentiation aligns with recent studies utilizing the experimental induction of RNT, which demonstrates the potential independence and distinct characteristics of both trait RNT and state RNT (Grant, Mills, Judah, & White, Reference Grant, Mills, Judah and White2021; LeMoult, Arditte, D’Avanzato, & Joormann, Reference LeMoult, Arditte, D’Avanzato and Joormann2013; Robinson & Alloy, Reference Robinson and Alloy2003; Wang, Song, Lee, & Zhang, Reference Wang, Song, Lee and Zhang2022). For example, Misaki et al. (Reference Misaki, Tsuchiyagaito, Guinjoan, Rohan and Paulus2023) highlighted that while RSFC alterations distinguish between healthy and depressed individuals, trait RNT in depressed individuals is more closely predicted by FC during an induced RNT scan rather than a resting-state scan, suggesting that RNT involves an active mental process not fully represented in the resting-state. While trait RNT measures an individual’s tendency to engage in RNT, induced RNT (capturing instant symptomatology) enables us to probe specific triggers, response patterns, and the phenomenological characteristics of RNT that are not captured by trait RNT alone. Thus, discerning the brain mechanisms that underlie both the trait and state aspects of RNT could have significant implications for clinical practice in terms of RNT remediation.
Given the results of our previous resting-state FC investigations and the prior literature, we aimed to further clarify the mechanistic basis of RNT by comparing insular FC during RNT induction with resting-state FC in individuals with major depressive disorder (MDD). Specifically, we employed a seed-to-whole-brain analysis using six insula subregions as seeds. We hypothesized that individuals with MDD would exhibit a more substantial increase in insular FC during RNT induction compared to resting state, with these alterations being more pronounced in the MDD individuals than in healthy controls (HCs). By investigating these neural dynamics, we seek to address the question: How do the FC patterns of the insula differ between resting state and RNT induction in MDD, and what implications do these differences have for the development of targeted neuromodulatory interventions?
Methods
Study design
The study protocol was reviewed and approved by the WCG IRB (https://www.wcgirb.com) (IRB Tracking No. 20210286) and registered on ClinicalTrials.gov (NCT04941066) as a part of a real-time fMRI-neurofeedback (rtfMRI-nf) study (Tsuchiyagaito et al., Reference Tsuchiyagaito, Misaki, Zoubi, Tulsa, Paulus and Bodurka2021, Reference Tsuchiyagaito, Misaki, Kirlic, Yu, Sanchez, Cochran, Stewart, Smith, Fitzgerald, Rohan, Paulus and Guinjoan2023b).
Participants
Forty-one MDD and 28 HC volunteers were recruited for rtfMRI-nf studies, making up a total of 69 participants (Tsuchiyagaito et al., Reference Tsuchiyagaito, Misaki, Zoubi, Tulsa, Paulus and Bodurka2021, Reference Tsuchiyagaito, Misaki, Kirlic, Yu, Sanchez, Cochran, Stewart, Smith, Fitzgerald, Rohan, Paulus and Guinjoan2023b). Participants were of both sexes, between the ages of 18 and 65 years, and fluent in English. Exclusion criteria were pregnancy, an abnormal neuromorphological brain profile as assessed by a radiology specialist physician, and other general contraindications for MRI safety. HC participants were defined based on the Mini-International Neuropsychiatric Interview 7.0.2 (MINI) (Sheehan et al., Reference Sheehan, Lecrubier, Sheehan, Amorim, Janavs, Weiller, Hergueta, Baker and Dunbar1998) and confirmed in a clinical conference with a board-certified psychiatrist. MDD-specific inclusion criteria included meeting the criteria of the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) for unipolar MDD based on the MINI (Sheehan et al., Reference Sheehan, Lecrubier, Sheehan, Amorim, Janavs, Weiller, Hergueta, Baker and Dunbar1998) and exhibiting current depressive symptoms with a Montgomery–Åsberg Depression Rating Scale (MADRS) score of >6 (Montgomery & Asberg, Reference Montgomery and Asberg1979). MDD-specific exclusion criteria were as follows: a lifetime history of bipolar disorder, schizophrenia, or any psychotic disorders; DSM-5 criteria for substance abuse or dependence within 6 months prior to study entry; active suicidal ideation as indicated by the Columbia-Suicide Severity Rating Scale (C-SSRS) (Posner et al., Reference Posner, Brown, Stanley, Brent, Yershova, Oquendo, Currier, Melvin, Greenhill, Shen and Mann2011) or an attempt within 12 months prior to study entry; commencement of psychotropic medication for depression and/or anxiety less than 1 month before the study enrollment; and commencement of psychological therapy less than 1 month before the study enrollment. All participants completed a written informed consent process before participating in the study.
Neuroimaging data acquisition
Neuroimaging was conducted on a 3 Tesla MR750 Discovery scanner (GE Healthcare, Milwaukee, WI) with an eight-channel, receive-only head array coil. Blood-oxygen-level-dependent fMRI data were acquired using a T2*-weighted gradient echo-planar sequence with sensitivity encoding (SENSE) with the following parameters: TR/TE = 2,000/25 ms; acquisition matrix = 96 × 96; FOV/slice = 240/2.9 mm; flip angle = 90°; voxel size = 2.5 × 2.5 × 2.9 mm3; 40 axial slices; and SENSE acceleration R = 2. To provide an anatomical reference for fMRI data, T1-weighted (T1w) MRI images were acquired with a magnetization-prepared rapid gradient-echo sequence with the parameters of FOV = 240 × 192 mm, matrix = 256 × 256, 124 axial slices, slice thickness = 1.2 mm, 0.94 × 0.94 × 1.2 mm3 voxel volume, TR/TE = 5/2 ms, SENSE acceleration R = 2, flip angle = 8°, delay/inversion time TD/TI = 1,400/725 ms, sampling bandwidth = 31.2 kHz, and scan time = 4 min 59 s.
Experimentally induced RNT and resting-state scanning
The MRI session started with a 5-min T1w MRI anatomical scan, a 6-min 50-s resting-state fMRI scan, and a 6-min 50-s experimentally induced RNT fMRI scan. Prior to the MRI session, participants identified a recent personal event that significantly triggered RNT, such as experiencing rejection by someone important to them. Participants provided a brief title for this event, which was used by research staff to prompt the participant’s recall immediately before the RNT-inducing fMRI scan. Participants were then instructed about the neurofeedback task as described in detail in Tsuchiyagaito et al. (Reference Tsuchiyagaito, Misaki, Zoubi, Tulsa, Paulus and Bodurka2021), (Reference Tsuchiyagaito, Misaki, Kirlic, Yu, Sanchez, Cochran, Stewart, Smith, Fitzgerald, Rohan, Paulus and Guinjoan2023b), and then had a rest period before the MRI session. In the scanner, the session began with a resting-state scan, where participants were instructed to clear their minds and not think of anything while viewing a fixation cross. This instruction was used intentionally to minimize engagement in spontaneous self-referential or ruminative thoughts, which could confound comparisons with the subsequent RNT-induction scan. After the resting-state scan, participants were reminded of their selected RNT event and instructed to focus on their emotional reactions to it and to reflect on why they responded the way they did – thus entering an RNT state during the scan. While keeping their gaze on the fixation cross, they were asked to focus on their emotional reactions to their chosen event and why they responded the way they did. This procedure aimed to engage the participants in a state of rumination and brooding, characteristic of RNT, while inside the scanner. The MRI session ended with neurofeedback scans as described elsewhere (Tsuchiyagaito et al., Reference Tsuchiyagaito, Misaki, Zoubi, Tulsa, Paulus and Bodurka2021, Reference Tsuchiyagaito, Misaki, Kirlic, Yu, Sanchez, Cochran, Stewart, Smith, Fitzgerald, Rohan, Paulus and Guinjoan2023b).
Symptom measures
Trait-RNT
The 22-item Ruminative Response Scale (RRS) (Nolen-Hoeksema & Morrow, Reference Nolen-Hoeksema and Morrow1991) was used to measure trait RNT. The RRS is composed of three subscales: the 5-item ‘brooding’ subscale (e.g. RRS-B item: think ‘why can’t I handle things better’), the 12-item ‘depressive rumination’ subscale (e.g. RRS-D item: think about all of your shortcomings, failings, faults, and mistakes), and the 5-item ‘reflection’ subscale (e.g. RRS-R item: write down what you are thinking and analyse it). It assesses an individual’s tendency or trait to ruminate when they feel sad or are faced with depressive symptoms. Participants are asked to indicate what they ‘generally do when feeling down, sad, or depressed’ using a 4-point Likert scale ranging from 1 (never) to 4 (always), representing the trait tendency. The items in the RRS-B measure how often people engage in RNT, the causes and consequences of RNT, or a passive comparison with unachieved goals – characteristics that are found to lead to worse prognoses of depression (Treynor, Gonzalez, & Nolen-Hoeksema, Reference Treynor, Gonzalez and Nolen-Hoeksema2003). The items in the RRS-D subscale are similar to the RRS-B subscale; however, this subscale measures how often people engage in RNT, with a focus on depressive symptoms and moods. We employed the RRS-B and RRS-D subscales for connectivity analyses related to trait RNT as RRS-R does not include pathological elements of RNT and may even reflect protective factors against depression (Treynor et al., Reference Treynor, Gonzalez and Nolen-Hoeksema2003).
State-RNT
The level of state RNT that immediately followed the state-RNT fMRI scan was assessed with the Visual Analogue Scale. Right after the resting-state and RNT-induction scans, participants used a button box to answer the question, ‘To what extent did you dwell on negative aspects of yourself?’. The answers consisted of ratings from 1 (not at all) to 10 (extremely), indicating the intensity of their state RNT during the scan.
Severity of depression and anxiety
Individuals with MDD were assessed before the MRI session using MADRS (Montgomery & Asberg, Reference Montgomery and Asberg1979) and the Hamilton Anxiety Scale (HAM-A) (Maier, Buller, Philipp, & Heuser, Reference Maier, Buller, Philipp and Heuser1988).
Preprocessing
Preprocessing of functional images was performed with Analysis of Functional NeuroImages (AFNI) (http://afni.nimh.nih.gov/afni/). The initial three volumes were excluded from the analysis. The preprocessing included despiking, RETROICOR (Glover, Li, & Ress, Reference Glover, Li and Ress2000), respiratory volume per time (Birn, Smith, Jones, & Bandettini, Reference Birn, Smith, Jones and Bandettini2008) physiological noise corrections, slice-timing correction, motion corrections, nonlinear warping to the MNI template brain with resampling to 2 mm3 voxels using the Advanced Normalization Tools (Avants, Epstein, Grossman, & Gee, Reference Avants, Epstein, Grossman and Gee2008) (http://stnava.github.io/ANTs/), smoothing with 6-mm FWHM kernel, and scaling to percent change relative to the mean signal in each voxel. We used FastSurfer (https://www.sciencedirect.com/science/article/pii/S1053811920304985) to extract white matter and ventricle masks from the anatomical image of an individual subject and then warped them to the normalized fMRI image space. General linear model (GLM) analysis was performed with regressors of 12 motion parameters (three rotations, three shifts, and their temporal derivatives), three principal components of ventricle signals, local white matter average signals (ANATICOR; Jo et al., Reference Jo, Saad, Simmons, Milbury and Cox2010), fourth-order Legendre polynomials for high-pass filtering, and censoring TRs with large head motion (>0.25-mm frame-wise displacement). Any data with more than 30% censored volumes were treated as a missing value for the group-level analysis (two datasets of HC during RNT induction and two datasets of MDD participants during RNT-induction and resting-state scans were treated as missing values). Voxel-wise residual signals of the GLM were used for the seed-to-whole brain analysis.
Seed-to-whole brain analysis
Definition of insular subregions
In order to better delineate the specific function of the insula, the Brainnetome insula sub-regions parcellation atlas was used (Fan et al., Reference Fan, Li, Zhuo, Zhang, Wang, Chen, Yang, Chu, Xie, Laird, Fox, Eickhoff, Yu and Jiang2016). This parcellation atlas defined fine-grained insular subregions using probabilistic connectivity patterns. The insula was segmented into six subregions in each hemisphere, including the hypergranular insula (G), ventral agranular insula (vla), dorsal agranular insula (Carney et al., Reference Carney, Blumenthal, Stein, Watkins, Catellier, Berkman, Czajkowski, O’Connor, Stone and Freedland2001; Sliz & Hayley, Reference Sliz and Hayley2012), ventral dysgranular and granular insula (vId/vIg), dorsal granular insula (dIg), and dorsal dysgranular insula (dId) (Supplementary Figure S1).
FC processing
Twelve seed-to-whole brain FC maps were calculated based on predefined insular subregions. The average time course was obtained from the seeds, and the FC maps were generated by calculating Pearson’s correlation coefficients between the time series within the seed and the time series from every other voxel across the whole brain. Correlation coefficients were converted to z-scores using Fisher’s r-to-z transformation.
Statistical analysis
AFNI’s 3dLMEr was performed on each seed to identify the connectivity patterns of the insular subregions with the interaction of diagnosis (MDD vs. HC) by run (RNT induction vs. Rest), age, sex, motion, and medication status as fixed effects, and subjects as random intercepts. Head motion was calculated using the six standard motion parameters (three translational and three rotational), and the mean framewise displacement for each scan was included as a covariate in the Linear Mixed Effects Modeling (LME modeling). The results of the main interaction effect, the main effect of diagnosis, and the main effect of run were reported as a chi-square statistic, and post hoc general linear t-style tests were specified in case of the significant main effect, as per the output of AFNI’s 3dLMEr. The significant threshold was set as peak p < 0.001 and cluster-wise p < 0.05/12 (Bonferroni-corrected). AFNI’s 3dClustSim with 10,000 permutation tests was employed to define the cluster-size thresholds (k > 143 voxels). Furthermore, exploratory correlation analyses were performed to investigate the association between changes in FC values during RNT-induction scans (relative to resting state), and both trait RNT and Δstate RNT (i.e. the difference between RNT induction and baseline resting state) in the MDD and HC groups, respectively. Average FC values were extracted from significant clusters identified in the main effect contrasts (RNT induction > Rest) for each seed region. These cluster-level values were then used in linear correlation analyses. An uncorrected threshold of p < 0.05 was applied, and the results were presented for exploratory purposes, as no correction for multiple comparisons was performed.
Results
Demographic and clinical measures
Table 1 shows the demographic data and clinical characteristics of the MDD (n = 41) and HC (n = 28) participants (total n = 69). The majority of these participants were Female and White, and over half of the MDD participants experienced anxiety disorder comorbidities (51.2%) and were treated with antidepressants (51.2%) (Table 1).
Demographic data
Abbreviations: HAM-A, Hamilton Anxiety Scale; MDRS, Montgomery–Åsberg Depression Rating Scale; MDD, major depressive disorder; RNT, repetitive negative thinking; RRS, Ruminative Response Scale; VAS, Visual Analogue Scale.
Insular-to-whole brain FC patterns
Interaction effect of diagnosis-by-run
We first examined the interaction effect of diagnosis (MDD vs. HC) by run (resting state vs. RNT induction). Contrary to our hypothesis, no significant FC alterations were observed for the diagnosis-by-run interaction across any of the insular subregions. Results with a threshold of p < 0.001, without cluster thresholding, are presented in Supplementary Figures S2 and S3.
Main effect of diagnosis and run
Participants with MDD demonstrated greater FC between the bilateral anterior, middle, and posterior insular regions and the cerebellum (z = 4.31–6.15). These results suggest a unique pattern of insular-cerebellar connectivity in MDD (Table 2 and Supplementary Figures S4 and S5).
Significant regions showing the main effect of diagnosis (MDD and HC) from seed-to-whole brain functional connectivity analysis
Abbreviations: HC, healthy control; MDD, major depressive disorder.
Regarding the main effect of run (Table 3 and Supplementary Figures S6 and S7), enhanced FC was found between the bilateral anterior and middle insula and other key brain regions, including the bilateral prefrontal cortices, parietal lobes, posterior cingulate cortex (PCC), and medial temporal gyrus, encompassing the STS (z = 4.47–8.31). Figure 1 displays additional spider charts and bar plots to illustrate the post hoc effects of these main findings.
Significant regions showing the main effect of run (RNT-induction and Rest) from seed-to-whole brain functional connectivity analysis
Abbreviations: PCC, posterior cingulate cortex; RNT, repetitive negative thinking; STS, superior temporal sulcus.
Post hoc investigation of (a) the effect of diagnosis (major depressive disorder vs. healthy control participants) and (b) the effect of run (RNT induction vs. Rest). Abbreviations: L, left; R, right; Cr, cerebellum; Vis, visual area; Ver, vermis; IFG, inferior frontal gyrus; OFG, orbital frontal gyrus; Ang, angular gyrus; PrCG, precentral gyrus; MTG, middle temporal gyrus; IPL, inferior parietal lobule; PCC, posterior cingulate cortex; SMG, supramarginal gyrus; SPL, superior parietal lobule; SMA, supplementary motor area; OpIFG, opercular part of the inferior frontal gyrus; RNT, repetitive negative thinking.
Correlation between insular FC and RNT measures
Figure 2 depicts significant associations between RNT measures and FC of the insular cortex with other regions in MDD participants, as well as HC participants. Consistent with our findings in increased insular FC during RNT induction relative to the resting state, among individuals with MDD, higher trait RNT was positively associated with increased FC between the right dorsal anterior and middle insula, the regions in the DMN (including the PCC and middle temporal gyrus), and the regions in the salience network (SN) (including the orbital frontal gyrus). Moreover, greater state-RNT scores during RNT induction, compared to resting-state, were positively correlated with increased FC in similar insular regions and the bilateral angular gyrus, as well as the right middle temporal gyrus (Figure 2). On the other hand, higher trait RNT was negatively correlated with increased insular FC between the left anterior insula and the inferior parietal lobule in individuals with MDD, although this FC showed an increased main effect of RNT induction (Table 3 and Figure 1).
Scatter plots and correlation between insular-cortical functional connectivity (FC) and RNT measures. (a) Correlation of trait RNT as measured by the Ruminative Response Scale-Brooding subscale (RRS-B) before the scan (x-axis) with changes in FC during RNT-induction scan compared to the Rest scan (y-axis). (b) Correlation of changes in state RNT as measured by the Visual Analogue Scale during RNT-induction scan compared to the Rest scan (x-axis) with changes in FC during RNT-induction scan compared to Rest scan (y-axis). Abbreviations: L, left; R, right; IPL, inferior parietal lobule; MTG, middle temporal gyrus; PCC, posterior cingulate cortex; SMG, supramarginal gyrus; Ang, angular gyrus; OFG, orbital frontal gyrus; SPL, superior parietal lobule; SMA, supplementary motor area; RNT, repetitive negative thinking.
Discussion
This study investigated the hypothesis that individuals with MDD would demonstrate a greater increase in FC between the insular cortex and other cortical (including cerebellar) regions during RNT induction compared to resting state. We also predicted that functional changes would be more pronounced in MDD, as compared with HC individuals. We observed three main findings during our research by which our hypothesis was partially supported. First, contrary to our hypothesis, there was no statistically significant diagnosis-by-run interaction in insular FC, indicating that changes in FC during RNT induction are not significantly different in individuals with MDD compared to HC individuals. Second, FC between insular and cerebellar cortices was higher in individuals with MDD compared to the HC group. Third, overall, FC between insular and other cortical regions increased during RNT induction compared to resting-state data.
Altogether, these findings support the hypothesis that the visceral control and higher-order cognitive processing changes underlie RNT intensity (Tsuchiyagaito et al., Reference Tsuchiyagaito, Sanchez, Misaki, Kuplicki, Park, Paulus and Guinjoan2022). These findings also reflect that insular-cortical FC was stronger during RNT induction compared with resting state. However, our results did not demonstrate a significant difference in FC alterations during RNT induction between the MDD and HC participants.
Insular connectivity in MDD
The observed higher FC between the anterior, middle, and posterior insula and the cerebellum in MDD participants aligns with emerging literature implicating cerebellar contributions to emotional and cognitive processes in depression (Depping, Schmitgen, Kubera, & Wolf, Reference Depping, Schmitgen, Kubera and Wolf2018; Habas, Reference Habas2021; Misaki et al., Reference Misaki, Tsuchiyagaito, Guinjoan, Rohan and Paulus2023; Pierce et al., Reference Pierce, Thomasson, Voruz, Selosse and Péron2023; Sliz & Hayley, Reference Sliz and Hayley2012; Van Overwalle et al., Reference Van Overwalle, Manto, Cattaneo, Clausi, Ferrari, Gabrieli, Guell, Heleven, Lupo, Ma, Michelutti, Olivito, Pu, Rice, Schmahmann, Siciliano, Sokolov, Stoodley, van Dun and Leggio2020). In our study, significant clusters were located in Crus I, Crus II, lobule VI, lobule VIII, and lobule IX, with bilateral involvement and some extension into the vermis. These regions have been associated with non-motor functions, including emotion regulation, self-referential processing, working memory, and affective appraisal (Buckner et al., Reference Buckner, Krienen, Castellanos, Diaz and Yeo2011; Depping et al., Reference Depping, Schmitgen, Kubera and Wolf2018; Stoodley & Schmahmann, Reference Stoodley and Schmahmann2009). For instance, Crus I and Crus II show intrinsic connectivity with prefrontal and parietal cortices and are considered part of a cognitive-affective cerebellar network (Depping et al., Reference Depping, Schmitgen, Kubera and Wolf2018; Pierce et al., Reference Pierce, Thomasson, Voruz, Selosse and Péron2023). Lobule VI serves as a transitional zone between motor and association regions, contributing to both sensorimotor control and emotional monitoring (Depping et al., Reference Depping, Schmitgen, Kubera and Wolf2018; Schmahmann, Reference Schmahmann2004). Additionally, lobule VIII and lobuleIX, though traditionally linked to sensorimotor functions, are increasingly recognized for their role in affective and visceral regulation, particularly in conjunction with the insula (Pierce et al., Reference Pierce, Thomasson, Voruz, Selosse and Péron2023). These findings suggest that increased insula–cerebellar FC observed in MDD may reflect dysregulation in circuits underlying interoceptive awareness, emotional salience detection, and self-focused negative appraisal, all of which are hallmarks of RNT (Depping et al., Reference Depping, Schmitgen, Kubera and Wolf2018; Habas, Reference Habas2021; Misaki et al., Reference Misaki, Tsuchiyagaito, Guinjoan, Rohan and Paulus2023; Pierce et al., Reference Pierce, Thomasson, Voruz, Selosse and Péron2023; Sliz & Hayley, Reference Sliz and Hayley2012; Van Overwalle et al., Reference Van Overwalle, Manto, Cattaneo, Clausi, Ferrari, Gabrieli, Guell, Heleven, Lupo, Ma, Michelutti, Olivito, Pu, Rice, Schmahmann, Siciliano, Sokolov, Stoodley, van Dun and Leggio2020). These findings support the notion that cerebellar subregions interact with cortical salience and DMNs during affective processing (Prati, Pontes-Silva, & Gianlorenço, Reference Prati, Pontes-Silva and Gianlorenço2024), and may contribute to the persistent and ruminative symptomatology of depression. Future studies employing cerebellar parcellation and task-based designs may further clarify the role of these distinct subregions in the pathophysiology of MDD.
The insula participates in a multitude of functions related to information processing and emotional feeling via its connections to the limbic and autonomic nervous systems (Gasquoine, Reference Gasquoine2014). In addition, prior neuroimaging studies have noted insula hyper-activation and increased blood flow and metabolism in patients with major depression and other mental disorders (Gasquoine, Reference Gasquoine2014). Given this, it is possible that the insula contributes to the negative symptomology experienced during RNT due to its role in interpreting interoceptive and perceptual data. Furthermore, known for integrating somatosensory, affective, and cognitive information (Sliz & Hayley, Reference Sliz and Hayley2012), the insula may be crucial in maintaining the heightened state of the negative self-focus aspect of RNT. Prior neuroimaging research has associated emotional recalling/remembrance and other cognitively demanding, emotional tasks with increased insular activity (Phan, Wager, Taylor, & Liberzon, Reference Phan, Wager, Taylor and Liberzon2002). In this context, our observation of increased FC between the insula and other brain regions during RNT-induction tasks is not surprising. Furthermore, this suggests that the trained regulation of insular activity, or decreasing FC between the bilateral insula and other areas such as the STS, the parietal cortices, or the PCC, may reduce state-RNT symptoms. For example, the emergence of focused deep brain neuromodulation or real-time fMRI neurofeedback in FC literature prompts deeper exploration of these brain activity regulation methods as a means to improve RNT in depression.
The cerebellum’s role in cognitive and emotional processing, traditionally recognized in the last two decades (Pierce et al., Reference Pierce, Thomasson, Voruz, Selosse and Péron2023; Rudolph et al., Reference Rudolph, Badura, Lutzu, Pathak, Thieme, Verpeut, Wagner, Yang and Fioravante2023), appears to be particularly significant in the context of mood disorders. Previous resting-state static and dynamic FC research identifies the cerebellum as having a vital role in emotional processing and executive functioning through its connections to the executive network, the DMN, the SN, the insula, and multiple brain cortical hubs associated with emotion regulation (Habas, Reference Habas2021). These connections suggest that the cerebellum is involved in frequent, multimodal collaboration with other crucial brain regions and networks to undertake the multifaceted nature of human emotions. Considering the cerebellum’s association with emotion regulation and other key networks, the cerebellum may also be an ROI worth investigating in future FC studies involving clinically depressed populations.
Alterations during RNT induction
The augmentation of FC during RNT induction between insular regions and areas, such as the prefrontal and parietal cortices, PCC, medial temporal gyrus, and the STS, is particularly noteworthy. These regions are implicated in a wide range of processes, from self-referential thought to emotional processing and memory retrieval. The increased connectivity that was noticed during RNT induction in our work suggests a heightened state of neural coordination in these networks, potentially underpinning the ruminative process. However, one key limitation to interpreting the relationship between RNT induction and the observed increase in insular connectivity stems from the absence of another specific comparator condition with which we could compare connectivity patterns among rest, RNT induction, and other instructed cognitive engagement (Pillet, Op de Beeck, & Lee Masson, Reference Pillet, Op de Beeck and Lee Masson2020). While the current findings highlight differences between RNT and rest, it is difficult to deduce strong conclusions about the neural activity that is unique to RNT and that which is distinctive of any other type of instructed cognitive activity.
Foregoing studies have presented similar results, demonstrating significant transformations in FC that occur in state RNT. In another mood-induction study, researchers found that increased connectivity between the DMN and the fronto-parietal network (FPN), along with decreased connectivity between the SN and the FPN, are both associated with increased RNT after experiencing sadness (Lydon-Staley et al., Reference Lydon-Staley, Kuehner, Zamoscik, Huffziger, Kirsch and Bassett2019). The changes in RNT-induced FC that were observed during our research, particularly with the MDD sample population, were congruent to the findings of their research. In these types of RNT-induction studies, a variety of key networks and brain regions can be observed at play in emotion regulation, many of which may serve as potential targets for interventions and future research aimed at reducing trait- and/or state-RNT symptoms.
Correlation between insular FC and trait- and state-RNT scores
The correlation of both trait- and state-RNT scores with increased FC in specific brain regions, particularly in MDD patients as reported herein, suggests that FC could be a potential biomarker for RNT severity in clinical settings. Specifically, trait-RNT scores were associated with the increased insular FC of several key regions in the DMN and orbitofrontal gyrus, which are implicated in self-referential and emotional processing (Northoff et al., Reference Northoff, Heinzel, De Greck, Bermpohl, Dobrowolny and Panksepp2006; Rempel-Clower, Reference Rempel-Clower2007). This association highlights the neural correlates of a general propensity to engage in RNT, reflecting a stable, trait-like aspect of cognitive processing in individuals. In contrast, state-RNT scores were associated with FC between the insula and the angular gyrus, as well as the right medial temporal gyrus, during experimentally induced RNT. Changes in state-RNT ratings indicate how participants engaged with RNT during experimental induction relative to the resting state. The association with increased FC in these regions suggests that the acute induction of RNT may engage neural circuits related to memory, conceptual processing (Deen et al., Reference Deen, Koldewyn, Kanwisher and Saxe2015; Humphreys, Ralph, & Simons, Reference Humphreys, Ralph and Simons2021; Ramanan, Piguet, & Irish, Reference Ramanan, Piguet and Irish2018; Seghier, Reference Seghier2013), and the integration of emotional and sensory information (Craig, Reference Craig2009). This distinction underlines the dynamic nature of RNT, where state-dependent increases in RNT were correlated with immediate neural responses, differentiating it from the more static trait RNT. Such findings illustrate the complex neural underpinnings of RNT, supporting the idea that different facets of RNT are potentially supported by different neural networks, as reported in prior studies (Rosenbaum et al., Reference Rosenbaum, Haipt, Fuhr, Haeussinger, Metzger, Nuerk, Fallgatter, Batra and Ehlis2017; Tsuchiyagaito et al., Reference Tsuchiyagaito, Misaki, Cochran, Philip, Paulus and Guinjoan2023a). However, we would caution against any definitive conclusions based on correlation analysis due to the exploratory nature of this analysis.
Limitations and future directions
While our findings contribute significantly to the understanding of RNT in MDD, several limitations must be acknowledged, including the small sample size and the absence of an additional specific comparator condition. Longitudinal studies, or interventional studies using emerging neuromodulation methods to noninvasively modulate the large-scale circuits described herein (Philip & Arulpragasam, Reference Philip and Arulpragasam2023), could help to establish a causative role of neural alterations in RNT. Additionally, incorporating another comparator condition (such as reflecting on a past positive or neutral experience) into the study protocol may help distinguish RNT-induced cognitive activity from resting or other instructed mental states, thereby improving interpretability (Pillet et al., Reference Pillet, Op de Beeck and Lee Masson2020). In the present study, we conducted separate linear mixed-effects models for each insular subregion to maintain analytic clarity and avoid overcomplicating the model structure. However, we acknowledge that an alternative approach, modeling all seed regions within a single LME framework that includes seed region, hemisphere, run, and diagnostic group as factors, could offer a more integrated view of insular connectivity. We did not pursue this strategy due to the analytic complexity of estimating higher-order interactions in a modestly sized sample, which could limit statistical power and affect model stability. Future studies with larger datasets may benefit from implementing such a comprehensive model to test for interaction effects across insular subregions in a unified manner.
Moreover, preceding research by our group and others has suggested that RNT is a transdiagnostic occurrence, as it is a usual feature in individuals with generalized anxiety disorder (GAD) and obsessive–compulsive disorder (OCD) (Wahl et al., Reference Wahl, Ehring, Kley, Lieb, Meyer, Kordon, Heinzel, Mazanec and Schönfeld2019). Given the comorbidity of these disorders, it may be worth conducting a similar investigation that explores FC developments and trait/state RNT with participants from GAD and OCD populations.
Conclusion
The findings of our study underscore the importance of insular connectivity in the neural systems underlying RNT in MDD. Individuals with MDD exhibit distinct FC patterns between the insula and the cerebellum, highlighting a neural circuit that may contribute to the persistence and intensity of RNT. In addition, both MDD and HC participants show increased insular connectivity with key brain regions, including the bilateral prefrontal cortices, parietal lobes, PCC, and medial temporal gyrus, during RNT induction compared to resting state. This suggests that the insula is part of a broader network that becomes more engaged during active RNT, facilitating the integration of emotional and cognitive aspects of negative self-related thoughts. However, it is again important to acknowledge that firm conclusions related to the differences in FC patterns noted between RNT induction and rest cannot be attributed uniquely to RNT due to the lack of an additional comparator condition.
Furthermore, higher trait-RNT in MDD participants was associated with increased connectivity between the insula and the regions within the DMN and the SN, indicating that persistent negative thinking is linked to specific insular connectivity patterns involving self-referential processing and emotional salience. These differential connectivity patterns, including regions where higher trait RNT is negatively correlated with increased insular connectivity, may serve as neural markers for the intensity of RNT. Taken together, our findings highlight an association between insular connectivity and its interactions with other brain regions in the manifestation of RNT, providing a foundation for the development of targeted neuromodulatory interventions to alleviate this symptom in depression. This is in line with emerging neuromodulation techniques with anatomical specificity (Mehić et al., Reference Mehić, Xu, Caler, Coulson, Moritz and Mourad2014; Siddiqi et al., Reference Siddiqi, Taylor, Cooke, Pascual-Leone, George and Fox2020) that can be used to modulate this circuitry.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291725100925.
Acknowledgements
We would like to express our appreciation to CoBRE NeuroMap Investigators at LIBR and all the research participants. We acknowledge the contributions of Sahib S. Khalsa, MD, PhD, Tim Collins, Dara Crittenden, Amy Peterson, Megan Cole, Lisa Kinyon, Lindsey Bailey, Courtney Boone, Natosha Markham, Lisa Rillo, Angela Yakshin, and the LIBR Assessment Team for diagnostic assessments and data collection, and Julie Arterbury, Leslie Walker, Amy Ginn, Bill Alden, Julie DiCarlo, and Greg Hammond for helping with MRI scanning. We also acknowledge Jerzy Bodurka, PhD (1964–2021) for his intellectual and scientific contributions to the establishment of the EEG, structural and functional MRI, and neurofeedback processes that provided the foundation for the data collection, analysis, and interpretation of findings for the present work.
Author contribution
Conceptualization: L.S.E. and A.T.; Formal analysis and methodology: L.S.E., M.M., and A.T.; Funding acquisition: M.P.P.; Resources: M.M. and M.P.P.; Supervision: M.P.P. and S.M.G.; Writing – original draft: L.S.E., S.M.G., and A.T.; Writing – review and editing: S.G., J.T., E.S.E., M.M., M.P.P., S.M.G., and E.J.W.
Funding statement
This work has been supported in part by the National Institute of General Medical Sciences Center (Grant No. P20GM121312) and the Laureate Institute for Brain Research. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Competing interests
M.P.P. is an advisor to Spring Care, Inc., a behavioral health startup, and he has received royalties for an article about methamphetamine in UpToDate. He has a consulting agreement with and receives compensation from F. Hoffmann-La Roche Ltd. The other authors report no financial relationships with commercial interests related to the present study.
Open access
