Neural Correlates of Rumination in Adolescents with Remitted Major Depressive Disorder and Healthy Controls

Abstract

The aim of the current study was to use fMRI to examine the neural correlates of engaging in rumination among a sample of remitted depressed adolescents, a population at high risk for future depressive relapse. A rumination induction task was used to assess differences in patterns of neural activation during rumination as compared with a distraction condition among 26 adolescents in remission from major depressive disorder (rMDD) and 15 healthy control adolescents. Self-report depression and rumination as well as clinician-rated depression were also assessed among all participants. All participants recruited regions in the default mode network (DMN), including the posterior cingulate cortex (PCC), medial prefrontal cortex (mPFC), inferior parietal lobe (IPL), and medial temporal gyrus (MTG) during rumination. Increased activation in these regions during rumination was correlated with increased self-report rumination and symptoms of depression across all participants. Adolescents with rMDD also exhibited greater activation in regions involved in visual, somatosensory, and emotion processing when compared to healthy peers. The current findings suggest that during ruminative thought, adolescents with rMDD are characterized by increased recruitment of regions within the DMN and in areas involved in visual, somatosensory, and emotion processing.

Major depressive disorder (MDD) during adolescence is associated with several debilitating outcomes, including a 5-fold increased risk for suicide attempts, increased occurrence of physical and medical hospitalizations, and impaired functioning in school, social, and family domains (e.g., Dunn et al., 2006). Adolescents with a history of MDD are also at a particularly high risk for depressive relapse, with a cumulative probability of recurrence of 40% by two years and 70% by five years (Avenevoli, Stolar, Li, Dierker, & Merikangas, 2001). Identifying risk factors that are present among adolescents during periods of remission from MDD may help shed light on the mechanisms implicated in depressive relapse.

One mechanism associated with risk for MDD relapse is rumination, a perseverative negative thought pattern characterized by repetitive, prolonged, self-reflective, and uncontrollable focus on sad mood and its causes and consequences (Nolen-Hoeksema, Larson, & Grayson, 1999). Research suggests that rumination is associated with greater severity and duration of depressive episodes in adults (for a review, see Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008), and prospectively increases the risk of depressive relapse (Roberts, Gilboa, & Gotlib, 1998; Marchetti, Koster, Sonuga-Barke, & Da-Raedt, 2012). Notably, adolescents with a past history of MDD also exhibit higher levels of self-report rumination relative to healthy controls (e.g., Beevers, Rohde, Stice, & Nolen-Hoeksema, 2007; Hong et al., 2010). Given the robust relation between extent of rumination and recurrent depressive course, identifying the neural mechanisms supporting rumination is essential in order to provide greater insight into how this mechanism may be modified to stave off treatment refractoriness.

To date, the most consistent neural correlates of rumination have been regions within the default mode network (DMN) and subgenual prefrontal cortex (sgPFC) (for a review, see Hamilton, Farmer, Fogelman, & Gotlib, 2015; Nejad, Fossati, Lemogne, 2013). The DMN includes regions such as the posterior cingulate cortex (PCC), ventro-medial prefrontal cortex (mPFC), medial temporal gyrus (MTG), and the inferior parietal cortex (IPC) and is activated during rest, mind-wandering, or self-reflective thought (for a review, see Whitfield-Gabrieli & Ford, 2012). Studies have suggested that adults with current major depressive disorder (MDD) exhibit greater activation in the subgenual ACC, PCC, mPFC, and MTG, encompassing the parahippocampus, when rumination is induced during functional magnetic resonance imaging (fMRI) (Cooney, Joormann, Eugene, Dennis, & Gotlib, 2010). Moreover, greater dominance of the DMN relative to a task positive network during resting state was associated with higher levels of depressive rumination in adults (Hamilton et al., 2011). Finally, meta-analytical findings demonstrate that increased functional connectivity between the DMN and sgPFC predicts levels of depressive rumination in adults (for a review, see Hamilton et al., 2015; Nejad, Fossati, Lemogne, 2013).

Laboratory studies have also shown that ruminative thought exacerbates negative mood among dysphoric and depressed individuals (e.g., Nolen-Hoeksema & Morrow, 1993). Given this, it is not surprising that rumination is also correlated with emotional processing networks, (for a review, see Nejad, Fossati, Lemogne, 2013). For example, self-report rumination is associated with amygdala activation when healthy controls are instructed to increase levels of negative affect (Ray et al., 2005). Several studies also suggest that rumination is correlated with greater sustained activation of the amygdala to emotional stimuli among depressed adults (Mandell, Siegle, Shutt, Feldmiller, & Thase, 2014; Siegle, Carter, & Thase, 2006; Siegle, Steinhauer, Thase, Stenger, & Carter, 2002). Moreover, this link has been replicated experimentally with depressed adults exhibiting greater activation in the amygdala while engaging in self-referential processing versus distraction (Cooney et al., 2010). Thus, regions supporting the encoding, processing, interpretation, and regulation of emotional information, such as the amygdala are also implicated in supporting ruminative processes, but have not yet been studied among adolescents.

Given that rumination is a known mechanism increasing risk for depressive relapse (Roberts et al., 1998), it is striking that no study to date has explored whether abnormalities in DMN and emotion processing regions during ruminative thought are present in adolescents with remitted major depressive disorder (rMDD). However, findings from neuroimaging studies exploring the neural correlates of emotion processing in adolescent depression may provide insight into regions that may support rumination and adolescent MDD. For instance, several studies highlight overactivation of the amygdala in response to processing negative stimuli in adolescents with current MDD (Beesdo et al., 2009; Yang et al., 2010) and among youth at high risk for MDD (Monk et al., 2008). Moreover, adolescents of depressed parents, a population at high risk for developing MDD, exhibited greater amygdala and prefrontal activation in response to a negative mood induction (Joormann, Cooney, Henry, & Gotlib, 2012). In addition to recruiting regions involved in emotional reactivity and regulation, Joormann and colleagues (2012) found that adolescents at high risk for MDD also recruited regions that are more characteristic of self-referential processing and rumination, such as the MTG during attempts to repair their negative mood. A critical, unanswered question is whether these regions may also be implicated in ruminative thought among adolescents with rMDD, a population at a particular high risk for relapse (Avenevoli, Stolar, Li, Dierker, & Merikangas, 2001).

The current study, therefore, sought to examine the neural correlates of engaging in ruminative thought versus distraction among youth in the remitted phase of depressive illness compared to their peers. The current task used to induce ruminative thought was adapted from previous experimental psychology studies (i.e., Cooney et al., 2010; Johnson et al., 2006; Lyubomirsky & Nolen-Hoeksema, 1993) and modified for an adolescent population. We hypothesized that adolescents with rMDD would demonstrate increased activation in the subgenual ACC and DMN regions, including activation in regions such as the PCC, IPC, and the mPFC relative to healthy control adolescents, during ruminative thought, relative to distraction. Consistent with previous studies examining the neural correlates of emotion processing in adolescent MDD (Beesdo et al., 2009; Yang et al., 2010), we also predicted that adolescents with rMDD would exhibit greater activation in the amygdala during ruminative thought, relative to distraction. Finally, to assess clinical relevance, exploratory correlations were run to examine the extent to which activations in regions associated with rumination versus distraction were associated with clinical ratings of depression and rumination.

Methods and Materials

Procedure

Participants were recruited using flyers, multiple forms of posting on the internet, and in-person at a pediatric psychiatry clinic. Patients with rMDD were recruited as part of a pilot clinical trial that involved randomization to 8 weeks of adjunctive treatment with Rumination-focused Cognitive Behavior Therapy (RFCBT; Watkins et al., 2011; Watkins, 2015) or assessment only. Healthy control (HC) participants were recruited for a baseline eligibility evaluation and one fMRI scan. The study was approved by the University of Illinois at Chicago s Institutional Review Board and all participants and one guardian signed assent and consent, respectively. Participants and their parents were then interviewed using the Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version (K-SADS-PL; Kaufman, Birmaher, Brent, & Rao, 1997) and the Children s Depression Rating Scale-Revised (CDRS-R; Poznanski & Mokros, 1996) by two trained research assistants supervised by a licensed psychologist. Adolescents also completed self-report measures including the Reynold s Adolescent Depression Scale (RADS; Reynolds, 1988), and the Ruminative Response Scale (RRS; Treynor, Gonzalez, & Nolen-Hoeksema, 2003). Within two weeks of this initial assessment, all participants underwent an fMRI scan.

Participants

Participants included adolescents in full or partial remission from MDD (n = 26) and adolescents with no lifetime history of any DSM-IV disorder (n = 15). Participants were considered in remission from MDD if they previously met criteria for at least one major depressive episode (MDE) but had not met full criteria for a MDE in the past two weeks. This was operationally defined as three or fewer full-threshold symptoms rated as clinically significant on the K-SADS-PL (Kaufman et al., 1997). Participants with more than three symptom criteria of depression at sub-threshold level (item rating of 2) on the K-SADS-PL were required to be enrolled in current mental health services. Of the 26 adolescents in remission from MDD, 2 subjects endorsed 1 current full-threshold depression symptom and 1 subject endorsed 3 current full-threshold depression symptoms on the K-SADS-PL (23 subjects endorsed zero full-threshold depression symptoms at the time of the study). As shown in Table 1, the average number of days since last MDE among adolescents in remission from MDD was 361.21. Participants were not eligible if they endorsed any signs of active suicidality or substance abuse or dependence within the previous six months. rMDD participants were permitted to be on a maintenance antidepressant or stimulant medication, but were only eligible if they had been on these medications for a minimum of 12 weeks with no dose changes in the last 2 weeks. HC participants could not meet current or past criteria for MDD or any other DSM-IV psychiatric disorder. The average age of our sample was 15.55 (SD = 1.78, Range = 12–18), 56% were Caucasian, and 56% were female. Participant demographics and clinical characteristics by group are presented in Table 1.

Table 1.

Demographic and Clinical Characteristics of rMDD and HC youth

	Remitted MDD (n = 26)		Healthy Controls (n = 15)
	Mean	SD	Mean	SD	t	p
Age	15.52	1.81	15.60	1.80	−0.14	0.89
WASI IQ Estimate	108.23	13.23	107.53	12.19	0.17	0.87
RADS	61.67	13.16	47.20	9.38	3.70	<0.01
CDRS-R	26.50	5.62	18.00	1.06	5.77	<0.001
RRS	52.40	12.07	30.43	7.11	6.21	<0.001
Rumination–Sadness Rating	2.47	1.09	2.29	.77	0.53	0.60
Rumination–Self-focus Rating	3.07	1.01	3.18	.49	−0.27	0.79
Distraction–Sadness Rating	1.78	.60	1.23	.47	2.83	<0.01
Distraction–Self-focus Rating	2.43	.67	2.60	.87	−0.63	0.54
Days Since Last Episode	361.21	570.28	-	-	-	-
	N	%	N	%	χ2	p
Gender (Female)	14	54.0	9	60.0	0.15	0.70
Race (Caucasian)	14	54.0	9	60.0	4.85	0.18
Psychiatric Medication	15	57.7	-	-	-	-
Current Comorbid Anxiety	10	24.4	-	-	-	-

Note: RADS = Reynolds Adolescent Depression Scale; CDRS-R = Children s Depression Rating Scale–Revised; RRS = Ruminative Response Scale; Sadness Rating = fMRI behavioral rating; Self-focus Rating = fMRI behavioral rating.

Clinical Measures

Children s Depression Rating Scale-Revised (CDRS-R; Poznanski & Mokros, 1996). The CDRS-R, a well validated 17-item clinician-rated depression severity measure based on interview of both the parent and child, was completed by an independent evaluator at baseline. The reliability and validity of the CDRS-R are well documented (Brooks & Kutcher, 2001). In this study, inter-rater reliability (intraclass correlations, ICC = 98%) on the CDRS-R was excellent.

Reynolds Adolescent Depression Scale (RADS; Reynolds, 1988). Adolescents at baseline completed the RADS, a 30-item self-report measure of current depressive symptoms. This measure is rated on a 4-point Likert scale and has excellent internal consistency and good retest reliability (Reynolds, 1988). Higher total scores reflected higher levels of depression. In the present study, α = .87.

Ruminative Response Scale (RRS; Treynor, Gonzalez, & Nolen-Hoeksema, 2003). Self-report rumination was assessed using the RRS. The RRS was used as a measure of depressive rumination given the sample s history of rumination and in order to compare the current findings to the existent literature among adults. The RRS consists of 22 possible responses to sad mood that are focused on the self, one s symptoms, and the possible causes and consequences of the mood state. The RRS had been shown to be a reliable and valid measure of rumination among young adolescent populations (Burwell & Shirk, 2007). In the present study, α = .94.

fMRI Stimulus Presentation

Rumination Paradigm. The fMRI rumination task was based on principles of previous studies (i.e., Cooney et al., 2010; Johnson et al., 2006; Lyubomirsky & Nolen-Hoeksema, 1993) and modified to include a negative mood induction in order to facilitate the emergence of ruminative thought and, therefore, decrease the amount of time the adolescent spent in the scanner. The task consisted of four blocks of a mood induction, rumination induction, and distraction induction. At the baseline visit, participants were asked to generate four negative life events (failure event, sad family event, hurtful event, and frustrating event). Approximately two weeks later, participants completed the fMRI scan. Following completion of a six-minute resting state scan (reported elsewhere), the rumination task began. The task proceeded with a mood induction instruction, rumination prompt, self-rating question, distraction prompt, and self-rating question. This order was repeated three additional times. The task was run in the same order with the same prompts for all participants. First, participants were given mood induction instructions for 25 seconds:’1) Remember the time when someone badly hurt your feelings; 2) Remember the saddest event in your family; 3) Remember the time you were so sad/frustrated that you felt there was no hope for you; or 4) Remember when you failed badly at something. For each of these memories, adolescents were asked to Use your imagination to bring this fully into your mind and picture the event. Immediately following each mood induction instruction, a rumination prompt was displayed for 30 seconds using Dr. Nolen-Hoeksema s standard rumination protocols and those adapted for a younger population (Hilt & Pollak, 2012) such as think about what your feelings mean. The rumination prompt was followed immediately by two self-report questions ( How sad do you feel right now? ) and ( How much are you focused on your feelings right now? ) on a 1 (“not at all”) to 4 (“a lot”) Likert scale; each question was displayed for 6 seconds for a total of 12 seconds. Next, a 10–15 second jittered crosshair interval was displayed followed by a 30 second distraction prompt. An example distraction prompt was to think about a truck full of watermelons. Task length was 8.5 minutes. After the scan was completed, a positive mood induction was used to remedy low mood. During the positive mood induction, participants were given detailed instructions to think of a positive memory and describe the memory in detail. Participants were then asked to rate their current mood; if low positive mood was endorsed, the clinical psychologist on call was notified to check in with the participant.

fMRI Acquisition and Preprocessing

The rumination/distraction task was completed in a 3.0 T GE Discovery scanner (Milwaukee, WI) using parallel imaging with ASSET and T2* gradient-echo axial EPI with the following parameters: 90 degree flip, field-of-view 22, matrix size = 64x64, slice thickness = 3mm, 22.2ms echo time, 44 slices, TR of 2000 ms, and a total of 265 volumes. Prior to the rumination/distraction task, an eyes-open resting scan was acquired over six minutes with the same parameters described above and a total of 240 volumes. High-resolution anatomic T1 spgr scans were obtained for spatial normalization. Motion was minimized with foam pads and a cross on the display, and by conveying the importance of holding still to participants. Several steps were taken to reduce potential sources of noise and artifact. Specifically, slice-timing correction was completed with SPM8 (http://www.fil.ion.ucl.ac.uk/spm/doc/) and motion detection algorithms (MCFLIRT, Jenkinson, Bannister, Brady, & Smith, 2002) were applied using FSL (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/). Coregistration of structural images to functional images was followed by spatial normalization of the coregistered T1-spgr to the Montreal Neurological Institute (MNI) template. The resulting normalization matrix was then applied to the slice-time-corrected, physiologically corrected, time series data. These normalized T2* time-series data were spatially smoothed with a 5mm Gaussian kernel resulting in T2* images with isotropic voxels, 2mm on a side.

Movement was addressed with the following strategies: first, blocks in which individual participants had any TR to TR movement exceeding 1.5mm, or 3 consecutive TRs exceeding the same in any direction - pitch, roll, or yaw were identified. In addition, normality plots of the average standard deviation of movement values were examined for outliers. Rumination and distraction blocks in which participants had movement values greater than two standard deviations beyond the mean were not modeled. This resulted in 3 participants having 1 rumination block omitted and 3 participants having 1 distraction block omitted from respective models. These 5 participants (3 rMDD and 2 HC) did not differ from participants without blocks omitted in terms of age, depressive severity, self-report rumination, sex, or depression diagnostic history (lowest p = .21). Further, the results described below remained unchanged when these participants were removed from analysis.

Consistent with previous research using a similar version of the rumination task in adults (Cooney et al. 2010; Johnson et al. 2006), contrast images were derived based upon the rumination versus distraction contrast. These were computed by using the standard hemodynamic response function in SPM to model Blood Oxygenation Level Dependent (BOLD) signal for all rumination blocks and subtracting similar BOLD signal changes for distraction blocks for each individual in a first level analysis. One-sample t-tests were conducted using whole brain analyses from the individual group contrasts between the rMDD and control groups in SPM8. Whole brain correction was achieved at p < .05 by using AlphaSim with 1000 Monte Carlo simulations to determine a joint threshold of height and extent (p < .005, cluster extent of 440 mm³). Post-hoc analyses included using a DMN mask which was derived from seed-based connectivity of the left PCC, a DMN hub, among the same sample of adolescents (Jacobs et al., under review). The DMN mask was entirely subsumed within an adult mask reported previously by Yeo et al., 2011. We created our own mask to be certain about developmental age relevance.

Behavioral Analyses

Participants completed ratings of state sadness and self-focus following each rumination and distraction block. Ratings were averaged across each condition (rumination, distraction), and a two-way (group: rMDD, HC) repeated condition (rumination, distraction) ANOVA was conducted to examine differences in sadness and self-focus.

Exploratory Factor Analysis

Exploratory factor analysis was conducted on extracted z values from each cluster of significant differences between groups (rMDD, HC) for the rumination versus the distraction contrast. The number of factors retained was determined using maximum likelihood as an extraction method and a threshold of eigenvalue >1, followed by oblique rotation. We also verified that extracted factors surpassed 50% of total variance. Bivariate correlations were then conducted examining the relation between extracted factors and self-report measures of depression and rumination. To rule out potential additional relationships with clinical and demographic features, we also examined whether participant s gender, medication status, and current anxiety comorbidity were related to the extracted factors during the rumination versus distraction contrast.

Results

Behavioral Results

When examining differences between the HC and rMDD groups in self-report sadness following the rumination and distraction blocks, results revealed significant main effects of group (F(1, 35) = 5.82, p = .02), reflecting that adolescents with rMDD reported higher levels of sadness across all four rumination and distraction prompts, and condition (F(1, 35) = 45.38, p < .001), reflecting greater sadness ratings across all participants during the rumination versus distraction condition, replicating previous adult studies using these inductions (Lyubomirsky & Nolen-Hoeksema, 1993), and confirming the manipulations were effective inside the fMRI scanner. The group × condition interaction was not significant (F(1, 35) = .29, p = .59). An identical ANOVA was conducted with self-focus ratings, and results revealed a significant main effect of condition, (F(1, 35) = 39.13, p < .001), reflecting that adolescents reported greater levels of focus on their feelings during the rumination versus distraction condition, again confirming the efficacy of the manipulations in the scanner. The main effect of group (F(1, 35) = .01, p = .96) and the group × condition interaction (F(1, 35) = 1.99, p = .17) were not significant.

fMRI Results

Rumination versus distraction across all participants

As shown in Table 2, across all participants, the rumination versus distraction contrast resulted in activations in default mode regions such as the posterior cingulate cortex (PCC; BA = 23), medial prefrontal cortex (mPFC; BA = 10), inferior parietal lobe (IPL; BA = 47), and medial temporal gyrus (MTG; BA=21). In addition, regions known to support the processing of emotional stimuli, such as the hippocampus, and visual stimuli, including the occipital gyrus (BA = 18) were activated during rumination compared to distraction. There were no statistically significant positive activations for the distraction versus rumination condition. Post-hoc analysis was conducted to determine if the regions activated during the rumination induction were inside or outside of the DMN using a mask created from the PCC seed of a resting state connectivity analysis within the same dataset (Jacobs et al., under review). Regions activated within and outside of the DMN during the rumination induction are presented in Figure 1.

Table 2.

Foci of significant activation for rumination versus distraction across participants.

Contrast/lobe	BA	MNI coordinates			Z	k	p-value
		x	y	z
Frontal
Inferior	47	36	22	−16	4.61	185	<.001
Medial	10	−6	54	4	4.91	524	<.001
Superior	6	−6	6	60	5.01	533	<.001
Temporal
Middle	21	−50	−32	−4	5.37	1045	<.001
Parietal
Precuneus	7	−4	−66	30	5.18	204	<.001
Limbic
Posterior Cingulate	23	−2	−46	22	5.74	514	<.001
Subcortical
Hippocampus/PHG		−20	−28	−6	3.59	97	<.001
Occipital
Cuneus	18	4	−84	26	4.03	153	<.001
Fusiform	18	−32	−72	−16	5.88	447	<.001
Middle	18	−18	−90	14	4.48	125	<.001
Lingual	18	14	−82	−14	6.05	372	<.001

Note: BA = Brodmann Areas; MINI = Montreal Neurological Institute; PHG = Parahippocampal Gyrus.

Figure 1.

Panel A. highlights the DMN network based upon bilateral posterior cingulate cortex (PCC), parahippocampal gyrus (HPF), and dorsomedial prefrontal cortex (dmPFC) rs-fMRI seeds (t > 15, k < 55) with our adolescent sample only (fuchsia), the DMN network from Yeo et al., 2011 only (cyan), and the intersection of the two masks (violet). Panel B, shows the activation for the contrast of rumination-distraction. Regions within the adolescent DMN mask are highlight in cyan, and regions in red include areas outside the adolescent DMN mask.

Group differences in rumination versus distraction

Adolescents in the rMDD group demonstrated greater activations in default mode regions including the left precuneus (BA = 7) and the right IPL (BA = 40) relative to HCs in the rumination-distraction contrast (Table 3). In addition, the left MTG (BA = 30) was activated to a greater extent in rMDD compared to HCs. Individuals with rMDD also recruited regions involved in visual processing such as the bilateral inferior occipital gyri (BA = 19), bilateral lingual gyri (BA = 19), and left fusiform (BA = 19) to a greater extent than HCs. Emotional processing regions including the amygdala, thalamus, and insula were also activated to a greater extent in rMDD compared to HCs during rumination compared to distraction. Finally, somatosensory areas, including bilateral precentral gyri (BAs = 4 and 44) were also hyperactive among rMDD individuals relative to HCs during rumination.¹ In contrast, there were no significant findings wherein HCs activated regions to a greater extent than rMDD youth in the rumination-distraction contrast.

Table 3.

Foci of significant activation for rumination versus distraction between groups.

Contrast/lobe	BA	MNI coordinates			Z	k	p- value	Factor	DMN
		x	y	z
rMDD > HC
Frontal
Precentral	44	56	6	10	3.88	544	<.001	1	Out
Precentral	4	−40	−10	44	3.73	1007	<.001	1	Out
Dorsal Cingulate	24	−2	4	44	3.85	1079	<.001	1	Out
Temporal
Fusiform	19	−40	−50	−14	4.73	996	<.001	2	Out
Middle	30	−32	−72	14	3.46	76	<.001	1	In
Temporal
Superior	22	64	−14	6	3.36	82	<.001	2	In
Parietal
Inferior	40	66	−34	34	3.71	474	<.001	2	In
Precuneus	7	−4	−42	56	3.99	1279	<.001	2	In
Precuneus	31	24	−68	30	3.49	715	<.001	2	In
Subcortical
Insula	13	40	−22	−2	3.82	279	<.001	1	Out
Thalamus/Putamen/Amygdala/PHG/SGACC	-	12	−12	0	3.91	2093	<.001	2	In/Out
Occipital
Inferior	19	46	−72	4	3.88	212	<.001	1	Out
Lingual	19	32	−58	4	4.00	600	<.001	1	Out
Lingual	19	−26	−70	2	3.25	69	<.005	2	Out

Note: BA = Brodmann Areas; DMN = Default Mode Network; MINI = Montreal Neurological Institute; PHG = Parahippocampal Gyrus. The last column denotes whether the significant region was inside or outside of the DMN mask.

We re-utilized the previously described DMN mask to examine whether regions activated during the rumination-distraction contrast among the rMDD adolescents were within or outside of the DMN. Apart from small overlaps, the majority of group differences during the rumination induction were outside of the DMN mask and included the previously described regions such as bilateral occipital gyri, thalamus, amygdala, insula, and bilateral precentral gyri.

Data Reduction with Exploratory Factor Analyses

Results from exploratory factor analysis suggested a two factor solution for activations in the rumination-distraction contrast for the rMDD adolescents. Factor 1 had an eigenvalue of 10.60 and accounted for 42% of the variance after rotation, whereas factor 2 has an eigenvalue of 1.30 and accounted for 37% of the variance after rotation. Factors 1 and 2 were positively related (r = .74, p < .001). As shown in Table 2, factor 1 had higher loadings on 8 of the 14 clusters, whereas factor 2 included 6 of the 14 clusters.

Clinical Correlates

As shown in Table 1, rMDD adolescents reported higher residual levels of rumination (t = 3.70, p < .01) as well as higher levels of self-report (t = 3.70, p < .001) and interviewer-rated (t = 5.77, p < .001) depressive symptoms, relative to HC adolescents. Across all adolescents, factor 1 during the rumination induction was associated with higher levels of rumination (r = .42, p < .01; Figure 3a) and interviewer-rated depressive symptoms (r = .32, p = .04); a non-significant trend emerged between factor 1 and self-report depressive symptoms (r = .28, p =.09). Factor 2 was also associated with higher levels of self-report rumination (r = .38, p =.02) and depressive symptoms (r = .32, p =.048), and greater clinician-rated depressive symptoms (r = .43, p < .01; Figure 3b).

Figure 3.

Relationships between Rumination-Distraction Activation and Rumination and Depression Symptoms.

Figure 3a. Rumination induction factor one is positively related to self-report rumination among adolescents.

Figure 3b. Rumination induction factor two is positively related to clinician-rated depressive symptoms among adolescents.

Potential Confounds

Finally, participant s gender, medication status, and current anxiety comorbidity were unrelated to the two extracted factors during the rumination versus distraction contrast (ps > .19). The rMDD adolescents continued to differ significantly from the HC adolescents across the two factors during the rumination versus distraction contrast when adjusting for behavioral ratings of sadness during the fMRI task (ps < .01), current anxiety symptoms (ps < .01), and after excluding rMDD participants currently on medication (ps < .04).

Discussion

The primary aim of the current study was to examine the neural correlates of rumination among youth in the remitted phase of depressive illness compared to healthy controls. The current study used an experimental task modified for fMRI that is known to induce rumination and distraction (Cooney et al. 2010), with slightly adapted language for use with adolescents. Across all participants, results indicated that adolescents reported greater levels of state sadness and self-focus while engaging in rumination relative to distraction, indicating that the experimental manipulation was successful in inducing ruminative mood. Reports of rumination were equivalent for both groups, whereas the depressed adolescents reported greater state sadness during the rumination induction relative to non-depressed adolescents. Engaging in rumination relative to distraction resulted in greater neural recruitment of regions in the DMN, including the PCC, mPFC, IPL, and MTG across all adolescents. Additional areas were also activated during the rumination induction among all adolescents, including areas involved in emotion processing, such as the hippocampus, as well as areas involved in visual processing. Beyond these main effects of condition that were predominantly within the DMN, rMDD adolescents exhibited greater activation than HC in a few DMN regions, including the precuneus, IPL, and MTG. In addition, the majority of group differences (rMDD > HC) during the rumination induction were outside of the DMN; and located in limbic, visual processing, and somatosensory areas. Notably, recruitment of these neural regions during rumination was also associated with greater levels of rumination and interviewer-rated depressive symptoms across participants. The current findings suggest that hyperactivation of these neural regions during rumination may be a residual consequence of adolescent-onset MDD or may be a risk factor for relapse. Given that rumination is a known predictor of depressive relapse (Roberts, Gilboa, & Gotlib, 1998; Marchetti et al., 2012), the current findings raise the possibility that these neural patterns may depict this risk.

The current results provide further support for the role of the DMN in rumination and extend previous findings by documenting this pattern in an adolescent population. Specifically, our finding that all adolescents exhibited greater DMN activation during ruminative thought relative to distraction is consistent with a previous study showing that adults exhibit greater activation of the PCC and precuneus during self-reflective thought (Johnson et al., 2006). The increased activation within regions of the DMN, such as the MTG is also consistent with previous findings from currently depressed adults (Cooney et al., 2011), and may suggest that adolescents in remission from MDD may be engaging in more mood-congruent memory retrieval during rumination when compared to healthy controls. Adolescents in remission from MDD recruited several areas outside of the DMN to a greater extent compared to healthy controls while engaging in rumination. Whereas links between emotional salience regions and rumination have been repeatedly observed in the actively depressed state (Cooney et al., 2011; Mandell et al., 2014; Siegle et al., 2006; Siegle et al., 2002), the current study is the first to illustrate heightened amygdala, insular, and thalamic activity during rumination that occurs during adolescence and persists during remission from a depressive episode.

It is also noteworthy that relative to healthy controls, adolescents in remission from MDD exhibited greater activation in visual processing and somatosensory areas during rumination relative to distraction, such as in the bilateral inferior occipital, bilateral lingual, left fusiform, and bilateral precentral gyri. Knowing that remitted depressed adolescents engage in higher levels of rumination relative to healthy controls (e.g., Beevers, Rohde, Stice, & Nolen-Hoeksema, 2007; Hong et al., 2010), rumination may be more elaborative for these individuals. In other words, adolescents in remission from depression may be more likely to ruminate on negative information more vividly with visual and somatic brain regions recruited. Importantly, these findings are consistent and extend upon a previous study highlighting the role of increased recruitment of visual processing areas, including the fusiform gyrus, in the retrieval of memories among remitted depressed adults, relative to healthy controls (van Wingen et al., 2010). The neural processes involved in rumination may parallel processes involved in the retrieval of negative, emotional memories and may be one mechanism placing these individuals at high risk for relapse.

Contrary to our hypothesis, the current study found no evidence for the role of the subgenual ACC being implicated in ruminative thought among remitted depressed adolescents at the uncorrected level. This is inconsistent with previous adult depression and rumination studies (Cooney et al., 2010; Grimm et al., 2009; Hamilton et al., 2015; Lemogne, Delaveau, Freton, Guionnet, & Fossati, 2012), although an uncorrected ROI analysis did show the expected effect. One possible explanation for this may be the focus on an adolescent population that is currently in remission from depression. That is, previous adult studies examining the neural correlates of ruminative thought in depression have not focused on individuals in remission from depression. Moreover, previous volumetric (Hasler, Drevets, Manji, Charney, 2004) and functional imaging (Matthews et al., 2009) studies suggest that abnormalities in the size and function of the subgenual ACC in depression are exacerbated by illness severity and chronicity. Thus, one possibility is that these regions are more likely to be implicated in rumination among adults with chronic MDD or multiple depressive episodes in the active state. Additionally, the current study did not examine connectivity between regions during rumination versus distraction. Given that previous studies consistently demonstrate that increased functional connectivity between the DMN and sgPFC is predictive of levels of depressive rumination in adults (for a review, see Hamilton et al., 2015; Nejad, Fossati, Lemogne, 2013), it is possible that connectivity analyses may reveal regions involved in rumination among adolescents in remission from MDD. Future studies are needed to test this possibility.

Interestingly, both healthy and rMDD adolescents appeared to engage in the rumination task to a similar degree, as reflected by their self-monitoring responses during the scan. This self-monitoring response assessed the extent to which adolescents engaged in rumination during short blocks in the scanner, which is a different phenomena than a tendency to ruminate as a response style in real-life (as measured by the RRS). Thus, consistent with other previous research (Beevers, Rohde, Stice, & Nolen-Hoeksema, 2007; Hong et al., 2010), these behavioral data suggest that rumination appears to be a trait-like factor implicated in adolescent MDD. The similar state-rumination responses in both groups might also suggest that rumination ratings of the self may have qualitatively different scales in MDD relative to HC. It is also noteworthy that the imaging results appear to be more robust and extensive than those observed in a similarly sized sample of adults with a similar task (Cooney et al., 2015), which could reflect developmental changes in capacity for and richness of rumination. Indeed, there are developmental considerations pertaining to rumination (Jose & Brown, 2008), as well as rumination during adolescent depression, which we were unable to address in this cross-sectional study. Therefore, it will be important for future, longitudinal studies to examine the stability and plasticity of these neural regions during ruminative thought from adolescence to adulthood.

There were additional limitations to the current study that provide important directions for future research. First, prospective designs are needed to determine if these neural abnormalities during rumination are present prior to the first onset of MDD, thereby representing a putative vulnerability factor. Second, the rMDD adolescents in the current sample were not medication naïve and had a high rate of psychiatric comorbidity. On the other hand, it is also possible that this may have resulted in a more generalizable sample given the high rates of comorbidity among adolescents with MDD (Costello, Mustillo, Erkanli, Keeler, & Angold, 2003; Kilpatrick et al., 2003). Third, the current study was preliminary in nature and included a small sample. Therefore, we were unable to examine the effect of age or gender. Given that past studies show that rumination helps to explain the gender difference in adolescent depression, (Jose & Brown, 2008), future studies with larger sample sizes are needed to examine whether gender moderates the relation between adolescent s depression history and neural activation during ruminative thought. Additionally, the current study was also unable to examine how neural activation during rumination correlated with different subscales of the RRS, as these subscales tend to be highly correlated among small samples. Thus, future research should examine the extent to which these neural regions during rumination are related to adaptive (self-reflective) versus maladaptive (brooding) forms of rumination among adolescents. Moreover, future studies should seek to discriminate the neural regions involved in mindwandering, background processing (e.g. to-do lists), distraction, and rumination, which are various forms of self-directed thought varying on degree to which they focus on concrete, abstract, and perseverative content (Watkins, 2008).

Finally, the task used in the current study to induce rumination had some limitations worth noting. First, the number of events in each condition was lower than that of typical fMRI studies in order to reduce the length of the experiment. In this initial study, we attempted to mirror adult studies utilizing a rumination induction. In subsequent studies, multiple control condition blocks might be employed to better segregate the unique pattern of activations associated with rumination. Probes can also be employed that provide qualitative information about experience of rumination (e.g., intensity, vividness). Second, in order to facilitate comparison with previous adult studies inducing rumination in the scanner (Cooney et al, 2010), the contrast used in the current study compared rumination with distraction. Thus, we cannot definitively conclude that regions differing between the rMDD and HC adolescents were specific to rumination, as they may reflect group differences in distraction, or order effects. Future studies including a baseline condition, such as a rest condition, or a divided attention task, are needed to parse regions specific to rumination versus distraction.

Finally, the current version of the rumination task included a negative mood induction in order to reduce the amount of time required to begin ruminating. A limitation of this design is that the neural regions activated in response to rumination cannot be discriminated from those deriving from the negative mood induction or memory of the events and conditions leading to rumination. Research suggests that girls at high risk for depression exhibit greater amygdala and prefrontal activation in response to a negative mood induction and increased MTG activation during attempts to repair their negative mood (Joormann et al., 2012). Data from the self-report validity questions administered immediately following each rumination and distraction block indicate that remitted depressed youth did not differ from healthy peers in their ratings of sadness following rumination. These data provide some assurance that the results obtained do not derive solely from the fact that the mood induction was more potent among the rMDD group. In addition, rumination may, in the regular course of daily life, be triggered by negative affect or stressful life events. Thus, this paradigm may provide a useful window into how that process affects those at risk for depression differently. Future research can include alternative rumination tasks to further explore the similar and distinct brain regions supporting sad mood versus rumination.

Conclusion

In summary, the current results suggest that hyperactivity of neural regions involved in the DMN, somatic and visual processing, and in the processing of emotional information is observed during ruminative thought among adolescents with rMDD relative to HC adolescents. Future studies are needed to examine the extent to which rMDD adolescents differ from healthy control youth in functional connectivity between neural regions involved in rumination, including regions of the DMN as well as the amygdala. Additionally, the use of more nuanced methods for examining networks, including connectomics will be an important next step to determine the neural networks involved in switching between rumination and distraction among adolescents, and to determine whether intervention efforts may be implemented to enhance the ability to disengage from rumination. The present examination of the neural signature of rumination using an fMRI task among a sample of adolescents during a period of relative wellness is innovative and important, and suggests that enhanced recruitment of neural regions involved in multisensory and self-referential processes while engaging in rumination may be a residual consequence of adolescent-onset MDD or a risk factor for relapse. Future studies should examine how targeting rumination through prevention or intervention may alter these neural mechanisms and, in turn, reduce risk for future depressive relapse.

Supplementary Material

13415_2016_486_MOESM1_ESM. Figure 1.

A posterior aspect of the subgenual ACC was present in the rMDD greater than HC during rumination versus distraction when using an uncorrected p-value. Panel A illustrates the whole brain contrast, and Panel B shows the BA 25 region with dilation of 2. Within the BA 25 mask from the Wake Forest Pickatlas, and an uncorrected threshold of p < .05 and k > 15, rMDD adolescents exhibited greater activation than HC during ruminative thought (Z = 3.55, p < .0002, k = 178, −4, 4, 10).

Figure 2.

Regions of significantly greater activation during rumination versus distraction in rMDD relative to HC.