Differential Patterns of Default Mode Network Activity Associated with Negative and Positive Rumination in Bipolar Disorder

Abstract

Background:

Patients with bipolar disorder (BD) engage in both negative and positive rumination, defined as maladaptive self-focused thinking, and this tendency predicts depressive and manic episodes, respectively. Prior research in patients with major depression implicates regions of the default mode network (DMN) consistent with the self-focused nature of rumination. Little is known about the neural correlates of rumination in bipolar disorder.

Methods:

Fifteen euthymic patients with BD (twelve with Type I) and 17 healthy controls (HC) performed negative and positive rumination induction tasks, as well as a distraction task, followed by a self-related trait judgment task while undergoing functional magnetic resonance imaging (fMRI). Participants also underwent resting state scans. We examined functional connectivity at rest and during the induction tasks, as well as task-based activation during the trait judgment task, in core regions of the DMN.

Results:

Compared to HC, patients with BD showed greater functional connectivity between the posterior cingulate cortex (PCC) and medial prefrontal cortex (MPFC) at rest and during positive rumination, compared to distraction. They also showed greater activity in the PCC and MPFC during processing of positive traits, following positive rumination. At rest and during negative rumination compared to distraction, patients with BD showed greater functional connectivity between the PCC and inferior parietal lobule than HC.

Conclusions:

These findings demonstrate that negative and positive rumination are subserved by different patterns of connectivity within the DMN in BD. Additionally, the PCC and MPFC are key regions involved in the processing of positive self-relevant traits following positive rumination.

Introduction

Reflecting on one’s past actions can provide much-needed insight to allow for personal growth and improvement. However, it can take a maladaptive form for patients with major depression (MDD) in the form of depressive or negative rumination, characterized by a passive repetitive negative self-focus and the possible causes and consequences of one’s depressed state. Since it was first conceptualized in the response styles theory as a response to low or depressed mood, the research on negative rumination has shown that it plays a central role in depression (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). The tendency to engage in negative rumination not only maintains depressed mood (Morrow & Nolen-Hoeksema, 1990; Nolen-Hoeksema, 1991) but predicts the likelihood of developing a major depressive episode (Nolen-Hoeksema & Morrow, 1991) and even an adverse clinical course (Kuehner & Weber, 1999). More recently, researchers have started to appreciate the role of negative rumination in other psychiatric disorders, most notably bipolar disorder (BD) (Ghaznavi & Deckersbach, 2012).

Rumination in Bipolar disorder

Bipolar disorder, which affects between 1–2.5% of the population (Merikangas et al., 2007), is characterized by hypomanic or manic episodes, often defined by elevated mood, with or without one or more episodes of depression. Similar to patients with major depression, patients with bipolar disorder also have a tendency to engage in negative rumination in response to low or depressed mood (Hanssen, Regeer, Schut, & Boelen, 2018; Johnson, McKenzie, & McMurrich, 2008) as well as between episodes of depression (Gruber, Eidelman, Johnson, Smith, & Harvey, 2011). This tendency to engage in negative rumination has also been found to be correlated with frequency of lifetime depression in patients with bipolar disorder (Gruber et al., 2011).

Patients with bipolar disorder also ruminate in response to an elevated or euphoric mood, albeit with a positive self-focus on positive affective experiences and qualities (Eisner, Johnson, & Carver, 2008; Hanssen et al., 2018; Johnson et al., 2008). This tendency to engage in positive rumination is present in patients with bipolar disorder, even in between episodes (Gruber et al., 2011). Additionally, positive rumination maintains and increases elevated mood (Eisner et al., 2008; Gruber, Harvey, & Johnson, 2009), is associated with lifetime frequency of mania (Gruber et al., 2011), and can even predict risk for mania (Eisner et al., 2008).

As is the case with hypomania or mania, which distinguishes bipolar disorder from major depressive disorder, so too does the tendency to engage in positive rumination; while both patient populations endorse a tendency to engage in negative rumination, only those with bipolar disorder endorse positive rumination, and this is greater when hypomanic/manic (Hanssen et al., 2018). This unique tendency to engage in positive rumination is true even when comparing depressed patients with BD and MDD; patients with bipolar disorder, who are currently depressed, still endorse a greater tendency to engage in positive rumination compared to patients with MDD (Weinstock, Chou, Celis-de Hoyos, Miller, & Gruber, 2018). One study did find that patients with MDD engage in positive rumination; however, they also found that the tendency to engage in positive rumination in those patients was correlated with elevated manic symptoms at 6 month follow-up (Gilbert, Nolen-Hoeksema, & Gruber, 2013). This likely reflects the subset of patients with MDD who have not yet had a hypomanic or manic episode, but who do go on to have one and are subsequently diagnosed with BD. If so, this suggests that the tendency to engage in positive rumination might even help identify those patients with BD who are initially incorrectly diagnosed with MDD.

Finally, there is some evidence to suggest that positive rumination may not only maintain and increase elevated mood but might also play a role in grandiose thinking (Bortolon & Raffard, 2021). Presumably, engaging in repetitive thinking with a positive self-focus leads to increased self-esteem and positive self-concept and, in turn, increased self-confidence and even grandiosity. In fact, people with high levels of manic vulnerability reported using more dampening of positive affect as a possible strategy to manage the emergence of manic symptoms (G. C. Feldman, J. Joormann, & S. L. Johnson, 2008). Consistent with this, patients with BD who reported more dampening responses to positive affect than control participants, have also reported avoiding at least one rewarding activity to prevent mania (Edge et al., 2013).

Of note, studies of negative and positive rumination in BD have either been conducted in samples including only patients with Type I BD (Edge et al., 2013; Gilbert et al., 2013; Grubert et al., 2009; Gruber et al., 2011; Weinstock et al., 2018), or do not address differences between patients with Type I and Type II BD in their mixed sample, with one exception. Hanssen et al (2018) examined differences in positive rumination between patients with Type I and Type II BD and did not find any differences in positive rumination, but did find that patients with Type II BD were more likely to engage in dampening in response to positive affect.

Neural Correlates of Rumination

Research on the neural correlates of rumination have implicated regions of the default mode network (DMN) (Buckner, Andrews-Hanna, & Schacter, 2008), which is known to subserve self-related processing (Northoff, 2007; Northoff et al., 2006). A recent meta-analysis of neuroimaging studies of rumination in healthy controls identified increased activation of the anterior medial prefrontal cortex (MPFC) and posterior cingulate cortex (PCC) as being hubs of the DMN during rumination (Zhou et al., 2020). Individuals with MDD have also been found to show higher levels of DMN dominance at rest, which is associated with higher levels of the tendency to engage in negative rumination (Hamilton et al., 2011). In one study, compared to healthy controls, patients with MDD who were induced to ruminate showed greater activation in areas of the DMN including the anterior cingulate cortex (ACC), PCC, MPFC, inferior parietal lobule (IPL), and parahippocampus (Cooney, Joormann, Eugène, Dennis, & Gotlib, 2010). A more recent study found that compared to healthy controls, individuals with MDD showed increased resting state functional connectivity in the MPFC/ventral anterior cingulate cortex and that this was positively correlated with the tendency to engage in negative rumination (Zhu et al., 2012).

With regards to the neural correlates of rumination in bipolar disorder, little is known. There is one study of euthymic patients with bipolar disorder, which showed that relative to healthy controls, there was greater activation in a network comprised of several midline brain areas usually associated with self-referential processing, including the MPFC, PCC, and precuneus when participants were directed to focus their attentions inwards on themselves as opposed to when they were directed to focus externally on the letters in a word (Apazoglou et al., 2019). Additionally, a higher tendency to engage in negative rumination was correlated with increased activity in the MPFC and PCC in the patients with bipolar disorder during self-directed focus.

Current Study

Given the central role that both negative and positive rumination play in the course of illness in bipolar disorder, and that positive rumination may be a distinguishing feature of bipolar disorder, research into the neural correlates of negative and positive rumination in bipolar disorder has the potential to identify targets for treatment as well as possible biomarkers for bipolar disorder. In the current study, we examined neural activation at rest and during a negative and positive rumination induction, as well as a distraction induction, in euthymic patients with bipolar disorder and healthy controls. Additionally, to explore the impact of rumination on self-related processing, and the possibility that positive rumination leads to grandiose self-concept, we examined neural activity on a self-related processing task following each of the induction conditions.

We predicted that compared to healthy controls, patients with bipolar disorder would show increased connectivity within DMN regions both at rest and during the rumination induction conditions. We also predicted that connectivity between the MPFC and PCC at rest and during the negative rumination condition would be correlated with the tendency to engage in negative rumination, and connectivity between the MPFC and PCC at rest and during the positive rumination induction would correlate with the tendency to engage in positive rumination. Finally, we predicted that DMN regions identified from the connectivity analyses would show increased activity in patients with bipolar disorder during the self-relevant processing task following the rumination induction conditions (e.g., during negative task stimuli after the negative rumination induction, and during positive task stimuli after the positive rumination induction).

Methods and Materials

Participants

Fifteen right-handed native English-speaking participants meeting DSM-V criteria for BD were recruited from the Dauten Family Center for Bipolar Treatment Innovation at Massachusetts General Hospital. The clinical diagnosis was made using the Mini International Neuropsychiatric Interview (MINI) administered by a trained clinician. All fifteen participants with BD (twelve met criteria for Type I BD and three met criteria for Type II BD) were relatively euthymic with Hamilton Depression Scale (HAM-D) score <12 and Young Mania Rating Scale (YMRS) score <8. Seventeen right-handed native English-speaking healthy control (HC) participants matched for age, and reading level, were recruited through advertisements and a hospital email listserv for healthy volunteers. There were scanner-related issues for one task for the first two HC, so an additional two HC were recruited to ensure adequate data for that task (see Supplement for details of inclusion and exclusion criteria for both groups). There were no significant differences in age, gender, education, or scores on the Wechsler Test of Adult Reading (WTAR) between groups (Table 1). This study was approved by the Mass General Brigham Institutional Review Board and all procedures were performed in accordance with relevant regulations. Written informed consent was obtained from all subjects prior to enrollment.

Table 1. Study Demographics.

Values are n (%) or mean ± SD and (range).

Variables	Healthy Controls (n = 17)	Patients with Bipolar Disorder (n = 15)	p-value
Age
Mean ± SD (Range)	32.24 ± 10.21 (21–56)	35.27 ± 9.66 (25–55)	0.41a
Gender (%)
Male	8 (47.1)	7 (46.7)	0.98b
Female	9 (52.9)	8 (53.3)
Other	0	0
Ethnicity (%)
Non-Hispanic	15 (88.2)	12 (80.0)	0.28b
Hispanic	2 (11.8)	1 (6.7)
No selected answer	0	2 (13.3)
Race/Origin (%)
White	13 (76.4)	10 (66.7)	0.46b
Black/African American	2 (11.8)	2 (13.3)
Asian	2 (11.8)	1 (6.7)
Native American/Alaskan Native	0	0
Native Hawaiian/Pacific Islander	0	0
Multiple races	0	0
No selected answer	0	2 (13.3)
Level of Education (%)
Less than High School	0	0	0.75b
High School/GED	1 (5.9)	0
College	3 (17.6)	3 (20.0)
Master’s	2 (11.8)	3 (20.0)
Doctorate	1 (5.9)	2 (13.3)
Other	1 (5.9)	0
No selected answer	9 (52.9)	7 (46.7)
WTAR Scores	117.0(13.34)	116.4 (8.28)	0.88a
Mood Stabilizers		Lithium 3
		Lithium and Lamotrigine 4
		Atypical Antipsychotic 2
		Atypical Antipsychotic and
		Lamotrigine 1
		Lamotrigine 6

^aSignificant differences between group demographics calculated using Student’s t-test.

^bSignificant differences between group demographics calculated using Pearson’s χ² test.

Resting State Scan

Participants were scanned with a resting state sequence during which they were instructed to keep their eyes open and look at a fixation cross presented in the center of the screen (See Supplement for details of imaging parameters).

Rumination Induction Tasks

Following the resting state scan, all participants completed three runs of two tasks, each run consisting of one of three induction tasks (negative rumination, positive rumination, distraction) and then a self-attribute judgment task (Figure 1). Each induction task consisted of 10 phrases, presented individually on the screen for 30 seconds each. Participants were instructed to think about and imagine the content of the statements on the screen. In the negative rumination induction task, the phrases consisted of a series of negative self-focused phrases. In the positive rumination induction task, the phrases were a series of positive self-focused phrases. Finally, in the distraction induction task, they were presented with a series of phrases not related to the self, directing attention externally from the self, and neutral in content. It is important to note that previous studies using rumination inductions generally used symptom and/or self-focused prompts which were neutral; however, given the goal of inducing rumination with a negative or positive valence, the prompts for this study were both self-focused and had a negative or positive valence. Please see Table 2 for prompts used for each induction. The order of presentation of the induction tasks was counterbalanced across participants.

Figure 1.

Participants completed three runs of two tasks, each run consisting of one of three induction tasks (negative rumination, positive rumination, distraction) and then a self-attribute judgment task. Order of presentation of induction tasks and self-attribute judgement tasks were counterbalanced across participants.

Table 2. Prompts used for each induction.

Participants were presented with the following prompts for each induction task.

Negative Rumination	Positive Rumination	Distraction
Think about an occasion where you failed at something important.	Think about a time when you achieved a very important goal.	Imagine a truckload of watermelons.
Think about a time when you let someone down.	Think about a time when you were helpful to someone.	Imagine rain drops on leaves.
Think about a time when you made a bad financial decision.	Think about a strength you are glad you have.	Imagine the pattern on an oriental rug.
Think about a weakness you wish you did not have.	Think about a time when you received praise.	Imagine sailboats on the water.
Think about a time when you did not get a job or were denied a promotion.	Think about a time when you felt productive.	Imagine the color of leaves in the fall.
Think about a time when you embarrassed yourself in front of other people.	Think about a time when you felt creative.	Imagine the pattern on a peacock.
Think about a relationship that failed and your role in its failing.	Think about a time when you felt excited.	Imagine a blue sky with clouds.
Think about a time when you felt hopeless.	Think about an occasion where you felt very accepted.	Imagine a city skyline.
Think about a time when you received negative criticism.	Think about a time when you felt you were entertaining.	Imagine a bowl of fruits.
Think about a time when you felt rejected.	Think about a time when you had lots of energy.	Imagine the pattern on a snowflake

Self-Attribution Task

In the self-attribution task, participants were presented with a series of trait adjectives individually on a screen and asked to make a judgment about whether that personality trait applies to them. They were asked to indicate “yes” or “no” using a button press. The words were a series of intermixed positively valenced (e.g. “loyal”), neutral (e.g. “quiet”), and negatively valenced (e.g. “dishonest”) attributes taken from Anderson (1968) (Anderson, 1968). Trait attributes were counterbalanced for word length, number of syllables, Kucera-Francis frequency, and Hyperspace Analogue to Language Frequency. The presentation of blocks of attributes following the induction tasks was counterbalanced across participants.

Clinical and Behavioral Measures

In addition to the HAM-D and YMRS to assess for mood symptoms, participants filled out the Ruminative Responses Scale (Nolen-Hoeksema, 1991), which assesses the tendency to engage in negative rumination in response to negative affect, i.e. low or depressed mood. Participants also completed the Responses to Positive Affect Scale (Greg C. Feldman, Jutta Joormann, & Sheri L. Johnson, 2008), which assesses the tendency to engage in positive rumination in response to positive affect, i.e. elevated or euphoric mood. (See supplement for details about measures used).

Imaging Acquisition

Participants were scanned using a 3T Siemens Skyra magnetic resonance imaging (MRI) scanner (Siemens Medical Solutions USA Inc., 51 Valley Stream Parkway Malvern, PA 19355 United States; (see Supplement for details).

Image Preprocessing Details

Functional data were preprocessed using SPM8 software (Wellcome Department of Cognitive Neurology, London, UK). For each individual subject, fMRI images within a time-series were first realigned to a reference image (the image in the middle of the time-series) using a least squares approach and a 6-parameter rigid body spatial transformation. This preprocessing step was done to remove motion-related artifacts and generated a set of realignment parameters. fMRI images were next coregistered. Images from the different scan sequences were realigned to each other. Segmentation was performed to produce grey matter, white matter, and cerebrospinal images for each subject’s structural scans. Each individual subject’s fMRI images were spatially/stereotactically normalized to the standardized normalized space established by the Montreal Neurological Institute (MNI; http://www.bic.mni.mcgill.ca). Finally, each individual subject’s fMRI images were smoothed/convolved with a three-dimensional Gaussian filter of 6 mm full-width at half maximum (FWHM). This was done to reduce noise due to residual differences in anatomy during group averaging. It should be noted that the default FWHM of the Gaussian smoothing kernel is 8 mm. Therefore, our Gaussian filter of 6 mm was more conservative.

Resting state functional connectivity analyses

Preprocessed images were entered into CONN toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012) (www.nitrc.org/projects/conn). Functional connectivity maps representing time-series correlations between blood oxygen level dependent (BOLD) signal from a previously defined PCC seed (Fox et al., 2005) and every voxel in the brain were estimated. BOLD signal from white matter and cerebrospinal fluid, and realignment parameters were identified as confounders and their effects were removed.

Second-level random effects analyses were then completed in SPM8. We completed ROI analyses at p < 0.005 with our a priori regions as defined by anatomical masks from the Wake Forest University Pick Atlas (Maldjian, Laurienti, Kraft, & Burdette, 2003). Our a priori regions of interest were based on core regions in the DMN defined by Buckner, Andrews-Hanna, and Schacter (2008) (Buckner et al., 2008): the medial prefrontal cortex (MPFC; Brodmann areas 9, 10, 24, 32), posterior cingulate/retrosplenial cortex (PCC/RSC; Brodmann areas 29, 30/23, 31), hippocampal formation (HF; bilateral hippocampus and hippocampal formation), lateral temporal cortex (LTC; Brodmann area 21), and inferior parietal lobule (IPL; Brodmann areas 39, 40). To control for multiple statistical comparisons, we maintained a cluster-level false positive detection rate at p < 0.05 with a cluster (k) extent empirically determined by Monte Carlo simulations in the AFNI program 3dClustSim (Cox, 1996; Forman et al., 1995).

We next extracted beta values from each subject’s first-level connectivity maps and conducted Pearson’s r bivariate correlations using SPSS Version 24 between these beta values and our behavioral measures.

Task-based analyses

For each subject, we investigated task-related differences by creating contrast images representing BOLD activation during each attribute type (i.e., negative, positive, and neutral) for each of the three runs of the task (i.e. following the positive rumination induction, the negative rumination induction, and distraction). We also explored the unique aspects of negative and positive attributes by creating individual contrasts for positive vs. neutral, negative vs. neutral, and positive vs. negative. These individual subject level contrast images were entered into a group second-level flexible factorial analysis in SPM8. The main contrasts of interest were BP > HC after positive rumination (positive > neutral), BP > HC after positive rumination (positive > negative). We also looked at the same contrasts for the task after negative rumination. We again completed ROI analyses at p < 0.05 with the same a priori regions in the DMN as mentioned above. To control for multiple statistical comparisons, we maintained a cluster-level false positive detection rate at p < 0.05 with a cluster (k) extent empirically determined by Monte Carlo simulations in the AFNI program 3dClustSim (Cox, 1996; Forman et al., 1995).

Results

Inferior Parietal Lobule and Negative Rumination

The BP group showed greater resting state functional connectivity between the PCC seed region and bilateral IPL compared with the HC group (right IPL t(29) = 4.40, p < 0.001, left IPL t(29) = 4.58, p < 0.001; Table 3, Figure 2A). The BP group reported higher Ruminative Responses Scale Total Scores than the HC group (t(29) = −5.561, p < 0.001) and IPL-PCC connectivity was correlated with Ruminative Responses Scale total scores (r(29) = 0.557, p = 0.002; Figure 2A). Individuals with BP also showed greater IPL-PCC connectivity than the HC group during a negative rumination induction block compared with a distraction block (F(1,29) = 12.08, p = 0.002; Table 3, Figure 2B).

Table 3. Significant Clusters in the DMN During Resting State, Inductions, and Task.

Cluster-level statistics for significant clusters in the default mode network (DMN) during resting state, inductions, and task. MNI = Montreal Neurological Institute

Region	Condition	Contrast	MNI Coordinates	k (voxels)	Z-Score
MPFC	Resting State	BP > HC	−10, 22, 44	91	3.48
			6, 30, 38	103	3.40
	Inductions	BP > HC Positive Rumination > Distraction	−54, 12, 40	77	3.01
	Me/Not Me Task	BP > HC After Positive Rumination Induction Positive > Neutral Attributes	6, 44, 26	203	2.80
			−6, 26, 18	168	3.15
		BP > HC After Positive Rumination Induction Positive > Negative Attributes	4, 48, 22	125	2.95
IPL	Resting State	BP > HC	60, −36, 40	503	3.85
			−38, −48, 44	467	3.29
	Inductions	BP > HC Negative Rumination > Distraction	−56, −44, 46	59	2.71
	Me/Not Me Task	Nothing Significant
LTC	Resting State	HC > BP	−52, 6, −22	53	2.91
	Inductions	BP > HC Negative Rumination > Positive Rumination	64, −52, 4	68	2.61
	Me/Not Me Task	Nothing Significant
HF	Resting State	Nothing Significant
	Inductions	BP > HC Positive Rumination > Distraction	30, −36, −8	77	4.03
		BP > HC Positive Rumination > Negative Rumination	−22, −14, −32	81	3.58
	Me/Not Me Task	Nothing Significant
PCC	Resting State	N/A (PCC was seed region)
	Inductions	Nothing Significant
	Me/Not Me Task	BP > HC After Positive Rumination Induction Positive > Neutral Attributes	−8, −42, 44	93	2.83
		BP > HC After Positive Rumination Induction Positive > Negative Attributes	−8, −12, 50	37	3.02

Figure 2.

Individuals with bipolar disorder (BP) showed A) greater resting state functional connectivity between the posterior cingulate cortex (PCC) seed region and the inferior parietal lobule (IPL) compared with healthy controls (HC). IPL-PCC connectivity was correlated with Ruminative Responses Scale total scores (r(29) = 0.557, p = 0.002). Individuals with BP also specifically showed B) greater IPL-PCC connectivity than the HC group during a negative rumination induction block compared with a distraction block. Overall, IPL connectivity and activation were associated with negative self-referential thinking in the BP group. Note: *p < 0.005.

Medial Prefrontal Cortex and Positive Rumination

Individuals with BP showed greater resting state functional connectivity between the PCC seed region and bilateral MPFC compared with the HC group (left MPFC t(29) = 4.49, p = 0.001, right MPFC t(29) = 4.49, p = 0.001; Table 3, Figure 3A). The BP group reported higher scores on the dampening subscale of the Response to Positive Affect Scale (RPA) than the HC group (t(29) = −3.074, p = 0.005). There were no significant differences between groups on the Emotion Focus and Self-Focus subscales of the RPA. Resting state MPFC-PCC connectivity was significantly correlated with Responses to Positive Affect Dampening subscale scores (r(30) = 0.361, p = 0.05; Figure 3A).

Figure 3.

Individuals with bipolar disorder (BP) showed A) greater resting state functional connectivity between the posterior cingulate cortex (PCC) seed region and the medial prefrontal cortex (MPFC) compared with healthy controls (HC). MPFC-PCC connectivity was correlated with Responses to Positive Affect Dampening subscale scores (r(30) = 0.361, p = 0.05). Individuals with BP also specifically showed B) greater MPFC-PCC connectivity than the HC group during a positive rumination induction block compared with a distraction block. Finally, after the positive rumination induction, individuals with BP showed C) greater MPFC activation specifically when making positive self-judgments compared with negative or neutral self-judgments relative to the HC group. Overall, MPFC connectivity and activation was associated with positive self-referential thinking in the BP group. Note: **p ≤ 0.001, *p < 0.05

Individuals with BP also specifically showed greater MPFC-PCC connectivity than the HC group during a positive rumination induction block compared with a distraction block (F(1,29) = 15.01, p = 0.001; Table 3, Figure 3B). Additionally, after the positive rumination induction, individuals with BP showed greater MPFC activation specifically when making positive self-judgments compared with negative (F(1,24) = 6.73, p = 0.016) or neutral self-judgments relative to the HC group (F(1,24) = 8.17, p = 0.009; Table 3, Figure 3C). Finally, activity in the MPFC for positive words following positive rumination was correlated with YMRS scores (r(25) = 0.410, p = 0.042).

Lateral Temporal Cortex

The HC group showed greater resting state functional connectivity between the PCC seed region and the LTC compared with individuals with BP (t(29) = −4.56, p < 0.001; Table 3, Figure 4A). LTC-PCC connectivity was negatively correlated with Ruminative Responses Scale total scores (r(30) = −0.628, p < 0.001; Figure 4A). Individuals with BP also specifically showed greater LTC-PCC connectivity than the HC group during a negative rumination induction block compared with a positive rumination induction block (F(1,29) = 11.13, p = 0.002; Table 3, Figure 4B).

Figure 4.

Healthy controls (HC) showed A) greater resting state functional connectivity between the posterior cingulate cortex (PCC) seed region and the lateral temporal cortex (LTC) compared with individuals with bipolar disorder (BP). LTC-PCC connectivity was negatively correlated with Ruminative Responses Scale total scores (r(30) = −0.628, p < 0.001). Individuals with BP also specifically showed B) greater LTC-PCC connectivity than the HC group during a negative rumination induction block compared with a positive induction block. Note: *p < 0.005

Hippocampal Formation

Individuals with BP specifically showed greater HF-PCC connectivity than the HC group during a positive rumination induction block compared with distraction (F(1,29) = 16.08, p < 0.001) and negative rumination blocks (F(1,29) = 10.50, p = 0.003; Table 3, Figure 5). Additionally, HF-PCC connectivity during the positive induction minus the HF-PCC connectivity during the distraction condition correlated with the RPA dampening subscale (r(30) = 0.378, p = 0.04), and this was trending for the HF-PCC connectivity during the positive induction minus the HF-PCC connectivity during the negative rumination condition (r(30) = 0.341, p = 0.065). There were no significant differences in HF at rest.

Figure 5.

Individuals with bipolar disorder (BP) specifically showed B) greater hippocampal formation (HF)-posterior cingulate cortex (PCC) connectivity than the healthy control (HC) group during a positive rumination induction block compared with distraction and negative rumination blocks. Note: **p < 0.001, *p < 0.005

Posterior Cingulate Cortex

After the positive rumination induction, relative to the HC group, individuals with BP showed greater activation in the PCC when making positive self-judgments compared to neutral (F(1,24) = 16.76, p < 0.001) and negative self-judgments (F(1,24) = 12.68, p = 0.002; Table 3, Figure 6).

Figure 6.

After the positive rumination induction, individuals with bipolar disorder (BP) showed greater activation in the posterior cingulate cortex (PCC) when making positive self-judgments compared to neutral and negative self-judgments, relative to healthy controls (HC). Note: **p < 0.001, *p < 0.005

Accounting for Group Heterogeneity

Given that our sample included both patients with Type I and Type II bipolar disorder, but a majority with Type I bipolar disorder (n =12), all analyses were carried out with just the patients with Type I bipolar disorder (please see supplement for detailed results), and our findings remained significant with one exception. The positive correlation between YMRS scores and mPFC activity for positive words following positive rumination was no longer significant with just patients with Type I bipolar disorder. Of note, patients with Type II bipolar disorder had higher YMRS scores than patients with Type I bipolar disorder (Type II mean = 5.50, SD = 0.707; Type I mean = 0.92, SD = 1.165, t(2.07) = −7.607, p = 0.015).

Discussion

Rumination, whether with a negative or positive focus, appears to play an important role in the etiology and maintenance of mood episodes in bipolar disorder. In the present study we investigated the neural correlates of negative and positive rumination in bipolar disorder, as well as their effects on self-relevant processing, to better understand the brain regions subserving both kinds of rumination and to explore how negative and positive rumination might contribute to mood episodes by affecting processing about the self.

Neural Correlates of Negative Rumination

Patients with bipolar disorder, compared to healthy controls, both at rest and when engaged in negative rumination showed greater IPL-PCC connectivity. What is more, IPL-PCC connectivity at rest was correlated with the tendency to engage in negative rumination. This is consistent with a previous finding that the IPL region of the DMN was found to be involved when patients with MDD were induced to ruminate (Cooney et al., 2010), presumably with a negative focus, given the tendency to engage in negative rumination in MDD. Interestingly the IPL is one of the brain regions which is believed to be involved in mind-wandering about the past, and in a recent study, participants who received cathodal transcranial direct current stimulation of the IPL, which is thought to have an inhibitory effect, showed a significant reduction in the frequency of negative mind-wandering thoughts about the past (Chou, Hooley, & Camprodon, 2020). Contrary to our prediction, we did not find a difference in activity in the IPL or PCC in patients with bipolar disorder when making a judgment about a negative personality trait compared to positive or neutral traits following the negative rumination induction.

The HC group showed greater resting state functional connectivity between the PCC seed region and the LTC compared with individuals with BP and this was negatively correlated with the tendency to engage in negative rumination. Previously, activity in the LTC has been associated with processing stimuli with lower self-relevance (Lou et al., 2004), so this increased connectivity in the HC group is consistent with lower levels of negative self-focused, i.e. negative ruminative, thought in the HC group.

Neural Correlates of Positive Rumination

We found that compared to healthy controls, patients with bipolar disorder, both at rest and when engaged in positive rumination, showed greater PCC-MPFC connectivity. We predicted that this connectivity would be correlated with the tendency to engage in positive rumination. However, given that there were no significant between group differences on the emotion focus and self-focus subscales of the RPA, which truly assess the tendency to engage in positive rumination, it is not surprising that we did not find a correlation between this connectivity at rest and the tendency to engage in positive rumination. Given the association between those two subscales of the RPA and hypomania or mania, it is possible that the combination of a largely euthymic group and small sample size left us underpowered to find meaningful information with regards to the tendency to engage in positive rumination. We did, however, find that the BP group reported higher scores on the dampening subscale of the RPA than healthy controls. Previously, research has shown that the dampening subscale, while correlated with vulnerability to mania, is not associated with current hypomanic/manic symptoms (Feldman et al., 2008). We did find that MPFC-PCC connectivity at rest was correlated with dampening subscale scores. This suggests that those who show this more robust connectivity associated with positive rumination might use dampening strategies to counter their increased risk for mania.

Additionally, as predicted, compared to HC, there was increased activity in the MPFC and PCC in patients with BD when making self-relevant judgments for positive traits following the positive rumination condition compared to neutral and negative traits. Furthermore, activity in the MPFC for positive words following positive rumination was correlated with current hypomanic symptoms as assessed by the YMRS. This suggests a relationship between current hypomanic symptoms, positive rumination, and positive self-judgment. Individuals who are already slightly hypomanic, who engage in positive rumination, activate the network consisting of the PCC-MPFC robustly, which in turn also leads to robust activity with positive self-attribution. This might be the neural mechanism by which positive rumination leads to inflated self-esteem and even grandiosity. If this is indeed the case, one way to prevent hypomania from evolving into mania, or even to treat mania, might be to target PCC-MPFC functional connectivity. Further research, especially prospective studies are needed to explore this possibility.

A Bias for the Positive

Notably, whereas negative rumination is seen in both MDD and BD, only positive rumination is seen in BD. In our study, we found that the neural correlates of a positive rumination were clearly connected to positive trait-judgements afterward, whereas neural correlates of negative rumination were not connected to negative trait-judgements. This is potentially indicative of what some have referred to as positive emotion persistence arising from a bias for the positive in bipolar disorder (Gruber, 2011). Consistent with this, we also found that individuals with BP showed greater HF-PCC connectivity than the HC group during the positive rumination induction compared with distraction and negative rumination. Given the role of the HF in autobiographical memory and the central role of autobiographical memories in rumination, this suggests a more robust involvement for autobiographical content with the positive rumination induction in the BP group than in the HC group compared with negative rumination, i.e., a bias for processing positive autobiographical content. This would explain why the effects on self-relevant processing of positive rumination were more significant than for negative rumination.

Limitations

Our study had certain important limitations, principle of which is our modest sample size. The sample was also a heterogenous sample, although our findings remained significant when we included only patients with Type I bipolar disorder, with one exception. The positive correlation between YMRS and activity in mPFC for positive words during the self-attribution task following positive rumination, though present, was no longer significant. This may reflect the fact that in our sample, patients with Type II BD had higher YMRS scores than patients with Type I BD, thus when we did not include patients with Type II BD, the range of YMRS scores became restricted. Additionally, there may have been an issue of reduced power with the removal of 3 patients, given that this was a three-way interaction (group x induction condition x trait valence). Another limitation of this study is that the patients were medicated, and with our data set, it is not possible to account for the effects of a given medication on functioning. Finally, we did not collect longitudinal data about which patients went on to develop hypomania/mania which would have informed an understanding about whether the patterns of functional connectivity during negative and positive rumination could predict the likelihood of a mood episodes.

Conclusion

In this study we investigated activity in the DMN at rest, during negative and positive rumination and during self-relevant processing in bipolar disorder. We observed relatively stable patterns of DMN activity across different levels of self-relevant processing: at rest, when participants were allowed to let their minds wander freely, when they were instructed to actively engage in positive or negative rumination, and when they were asked to actively consider the self-relevance of positive and negative attributes. Previous work in mood disorders have either principally focused on resting state activity or during self-relevant processing alone.

In conclusion, we demonstrate that negative and positive rumination are subserved by different functional connectivity within the DMN in patients with bipolar disorder, which further clarifies the role of different DMN regions in self-relevant processing. Whereas IPL-PCC connectivity was associated the tendency to engage in negative rumination and ruminating with a negative self-focus, which is also the case in MDD, MPFC-PCC connectivity was associated with ruminating with a positive self-focus. Additionally, the MPFC and PCC are central to processing positive self-relevant stimuli following positive rumination in bipolar disorder, which presents a promising avenue of further exploration about the possible neural mechanisms underlying how positive rumination leads to mania. Lastly, this study highlights the importance of investigating positive rumination as both a potential biomarker for BD as well as a potential target for the prevention and treatment of mania with neuromodulation.

Highlights.

• Negative and positive rumination in patients with bipolar disorder.

• Self-relevant processing at rest, during rumination, and an attribute judgment task.

• Default mode network in self-relevant processing in bipolar disorder.