مدولاسیون فعالیت فرونتو-لیمبیک توسط آموزش روانی در بیماران دو قطبی خلق طبیعی: یک مطالعه کاربردی MRI

Abstract Bipolar disorders (BD) are mainly characterized by emotional and cognitive processing impairment. The cerebral substrate explaining BD impairment and the action mechanisms of therapies are not completely understood, especially for psychosocial interventions. This fMRI study aims at assessing cerebral correlates of euthymic bipolar patients (EBP) before and after psychoeducation therapy. Sixteen EBP and 16 matched healthy subjects (HS) performed a word-face emotional Stroop task in two separate fMRI sessions at 3-month interval. Between fMRI sessions, EBP underwent psychoeducation. Before psychoeducation, the comparison of EBP vs. HS in fMRI data revealed (a) significant decreased activity of cognitive control regions such as bilateral inferior and left superior frontal gyri, right insula, right fusiform gyrus and bilateral occipital gyri and (b) significant increased activity of emotion-related processing regions such as bilateral hippocampus, parahippocampal gyri and the left middle temporal gyrus. After psychoeducation, EBP showed significant clinical improvement, increased activity of inferior frontal gyri and a tendency toward decreased activity of right hippocampus and parahippocampal gyrus. These results suggest that the imbalance between cognitive control and emotion processing systems characterizing BD acute episodes may persist during euthymic periods. Moreover, this imbalance may be improved by psychoeducation, which enhances the cognitive control and modulates emotional fluctuations in EBP.

1. Introduction Bipolar disorders (BD) are characterized by abnormal emotional and cognitive processing during thymic critical episodes (mania and depression) and inter-critical (euthymic) periods (Soreca et al., 2009 and Leboyer and Kupfer, 2010). Thymic episodes are associated with socio-professional disability, deterioration in the illness’ evolution, increased risk of comorbidities, addictions, and suicide (Goldstein et al., 2011 and Undurraga et al., 2011). Inter-critical periods are nevertheless not free of residual cognitive and/or emotional symptoms (Judd et al., 2005, Torrent et al., 2006 and Bauer et al., 2010). Some of these symptoms actually represent trait characteristics of the illness. The identification of trait abnormalities in BD and the related cerebral substrate is particularly important as it may contribute to early diagnosis of BD thus reducing the latency to adequate treatment and improving outcome ( Cusin et al., 2000 and Altamura et al., 2010). Current neurophysiological models suggest that mood dysregulation in BD may be explained by the imbalance between the limbic and prefrontal cerebral networks: first, an overactivation of both limbic and prefrontal regions involved in emotional perception and identification. It includes subcortical (ventral striatum, thalamus) and medial temporal (amygdala, hippocampus and parahippocampal gyrus) regions, and also some prefrontal regions such as the orbitofrontal cortex (OFC) and the rostral part of the anterior cingulate cortex (rACC). Second, an hypoactivation of prefrontal areas responsible for executive functions, attention and emotion regulation, including the dorsolateral prefrontal cortex (dlPFC), ventrolateral prefrontal cortex (vlPFC) and the dorsal part of the anterior cingulate cortex (dACC) (Phillips et al., 2003, Phillips et al., 2008a, Strakowski et al., 2005b and Strakowski et al., 2012). However, it has not been clearly demonstrated whether these fronto-limbic abnormalities persist during euthymic states, which would constitute trait abnormalities of BD (Hariri, 2012). Previous functional magnetic resonance imaging (fMRI) studies that compared cerebral activity of euthymic bipolar patients (EBP) to healthy subjects (HS) usually used two broad classes of activation paradigms: emotional or cognitive ( Chen et al., 2011). With respect to emotional paradigms, the majority of previous studies used emotional facial expressions under various tasks. Some of them exhibited increased limbic activity in EBP compared to HS by using stimuli varying in emotional intensity. Specifically, EBP vs. HS showed increased activity of left striatum in response to mild happy faces ( Hassel et al., 2008), of the left amygdala and left hippocampus for emotional vs. neutral faces processing ( Chen et al., 2010), and of the left putamen in response to mild fearful faces ( Surguladze et al., 2010). However, decreased limbic activity was also showed in EBP compared to HS during emotional tasks. Particularly, decreased activity was found within bilateral amygdala and temporal pole by using backward masking paradigm ( Van der Schot et al., 2010) and within ventral ACC, OFC and striatum in response to happy and neutral faces ( Liu et al., 2012). The use of a face matching task did not reveal significant amygdalar activity between EBP and HS ( Robinson et al., 2008). Considering the prefrontal cortex (PFC), the use of a face matching task ( Robinson et al., 2008) increased the activity of the right inferior frontal gyrus (IFG) in EBP compared to HS. Moreover, in response to fearful and happy faces the medial PFC showed increased activity in EBP ( Surguladze et al., 2010). Furthermore, compared to HS, the EBP showed decreased activation of the right IFG in the labeling emotion condition of a face matching task ( Foland-Ross et al., 2012), decreased activation of the left IFG and the left middle PFC in response to facial expression of disgust ( Malhi et al., 2007b), decreased activation of the right dorsolateral PFC (dlPFC) in response to neutral, mild and intense happy faces as well as within the left dlPFC in response to neutral, mild and intense fearful faces ( Hassel et al., 2008). Similarly, previous fMRI studies testing cognitive control processes showed increased IFG activity in EBP compared to HS during a color-word Stroop task (Blumberg et al., 2003) and a Continuous Performance Task (CPT) (Strakowski et al., 2004). Other studies found decreased right middle PFC activation in EBP compared to HS by using a counting Stroop interference task (Strakowski et al., 2005a) and decreased activity of left IFG and dlPFC during a color-word Stroop task (Kronhaus et al., 2006). In respect with limbic regions, increased activation of left parahippocampal/amygdala during a CPT (Strakowski et al., 2004) and decreased activation of left fronto-polar cortex and bilateral amygdala during a Go/Nogo task (Kaladjian et al., 2009a) have been shown in EBP compared to HS. These inconsistencies may be due to several factors such as medication effect or comorbidities (Phillips et al., 2008b), but the variability of task paradigms used may probably be the most important factor. Indeed, most fMRI studies used either cognitive or emotional tasks but few of them employed tasks that involved both cognitive and emotion processes that may better approach the emotion regulation processes (Malhi et al., 2005, Malhi et al., 2007a, Lagopoulos and Malhi, 2007 and Wessa et al., 2007). However, using an emotional Stroop task, both increased (Lagopoulos and Malhi, 2007) and decreased (Malhi et al., 2005) activations of the limbic system were shown in EBP compared to HS. With an emotional go/nogo task an increased overall activation of the fronto-striatal network in EBP was reported (Wessa et al., 2007). Consequently, it is necessary henceforth to more precisely identify trait characteristics of the BD using tasks designed to involve both emotional and cognitive processing that may also assess emotion regulation processes in BD patients (Phillips et al., 2008a). An emotional word-face Stroop adapted from Etkin et al. (2006) was used in the current study. By using emotional words and faces that may be congruent or not, this task permits to implicitly distract attentional control by emotional material and therefore examine neural systems involved in emotional processing and cognitive control interaction. Because this task included fearful, happy and neutral facial stimuli, it also allows assessing the impact of the emotional valence and arousal dimensions in emotional processing. Furthermore, next to a better understanding of the pathophysiology of BD, the development of therapeutic strategies constitutes a real challenge. Indeed, despite relative effectiveness in a majority of patients, pharmacological treatments are insufficient on a functional level, as well as on residual depressive, dysthymic and dysphoric symptoms (Tohen et al., 2000 and Huxley and Baldessarini, 2007). Consequently, in parallel with pharmacological progress, psychosocial interventions have recently undergone great development (Swartz and Frank, 2001, Zaretsky, 2003, Colom and Vieta, 2004 and Miklowitz, 2008). Most of these interventions target emotional and cognitive processes in order to improve adaptive processes and reach functional recovery (Honig et al., 1997 and Bernhard et al., 2006). Among various therapeutic approaches, clinicians, therapists and researchers have recently shown a particular interest in psychoeducation for BD treatment (Colom et al., 2003a and Rouget and Aubry, 2007). The aim of this approach is to teach patients to better manage BD symptoms in everyday life, to improve coping strategies and to optimize compliance with pharmacological treatment in order to prevent thymic relapses and improve functioning (Perry et al., 1999 and Colom et al., 2003b). Positive outcomes of psychoeducation in BD have been observed rapidly and are long lasting (i.e., five years), particularly in terms of risk, duration and severity of relapses (Colom et al., 2003a, Colom et al., 2009 and Rouget and Aubry, 2007). Moreover, positive effects have been observed in terms of quality of life (Michalak et al., 2005) and social functioning (Perry et al., 1999). Psychoeducation benefits in BD are similar to those revealed by CBT (Miklowitz, 2008 and Costa et al., 2010), which are nevertheless more time- and care- consuming to allow detection positive effects (Zaretsky et al., 2008). Despite significant improvements in clinical symptoms, the behavioral and neural mechanisms associated with psychoeducation are not completely understood (Miklowitz and Scott, 2009). It has been suggested that the mechanism of psychotherapeutic action would be top-down ( Mayberg et al., 1999) as it first involves modulation of cortical activity with a subsequently impact on subcortical regions. In contrast, pharmacological treatments could act in a bottom-up way ( Mayberg et al., 1999 and Mayberg, 2009) as they act first on a subcortical level (neurotransmitters brain centers), then modulating activity at a higher cortical level. The present study compares EBP and HS cerebral activity during performance of a task involving both cognitive and emotional processes, a word-face emotional Stroop (Etkin et al., 2006). Two objectives have been defined. First, we aimed to better identify neurofunctional abnormalities in EBP that could be the trait characteristics of BD. At the behavioral level, we assumed an increased emotional interference in EBP compared to HS, which may result in lower task performances in emotionally incongruent condition compared to congruent condition. At the cerebral level, EBP would reveal decreased activity in prefrontal regions during emotional conflict processing and increased activation of limbic regions during emotional processing. Second, we aimed to assess the effect of psychoeducation at the cerebral level. In order to answer this question, we compared EBP cerebral activity before and after psychoeducation. We assumed that psychoeducation modulates the activity of prefrontal and limbic networks underlying cognitive control and generation of emotional responses, respectively.

3. Results 3.1. Results EBP vs. HS at t1 3.1.1. Task performance ANOVAs revealed significant effect of Stroop condition on both RT and %CR [F(1,26)=11.91, P<0.001; F(1,26)=29.61, P<0.001, respectively] indicating robust behavioral interference associated with emotional conflict in both groups. The valence main effect was also significant on both RT and %CR [F(1,26)=44.83, P<0.001; F(1,26)=4.62, P=0.04, respectively]. Negative stimuli were processed slower and less correctly than positive ones in both groups. In addition, significant Stroop by Valence interaction was observed on RT [F(1,26)=4.54, P=0.04] and on %CR [F(1,26)=8.17, P=0.008]. Planned comparisons revealed that incongruent negative (IN) stimuli generated slower RT [F(1,26)=27.67, P<0.001] and more errors [F(1,26)=6.20, P=0.02] than incongruent positive (IP) stimuli. Furthermore, compared to HS, EBP were slower irrespective of Stroop and Valence conditions [F(1,26)=11.91, P=0.002], but they showed similar %CR. This result indicates that in EBP, correct response accuracy, similar to HS, was obtained with higher time cost (response slowing). We did not observe any significant Group×Stroop and Group×Valence interactions neither on RT [F(1,26)=0.02, P=0.89; F(1,26)=0.16, P=0.69, respectively] nor on % CR [F(1,26)=0.11, P=0.75; F(1,26)=1.35, P=0.26] (see Table 2 for descriptive data). Table 2. Behavioral performances for word-face emotional Stroop measured during fMRI at t1 and during fMRI at t2 in all participants. Note: Data are reported as mean (SD). HS=Healthy Subjects; EBP=Euthymic Bipolar Patients. Congruent positive Congruent negative Incongruent positive Incongruent negative Response time (ms) HS (t1) 770.8 (87.5) 891.0 (118.0) 871.0 (129.2) 983.3 (165.7) HS (t2) 737.1 (74.5) 852.6 (104.3) 820.6 (89.8) 934.9 (152.9) EBP (t1) 945.3 (228.9) 1060.2 (363.7) 1073.1 (295.9) 1129.4 (260.4) EBP (t2) 887.1 (87.5) 1038.2 (159.1) 1006.2 (123.6) 1093.6 (143.5) % Correct responses HS (t1) 98.2 (3.1) 93.4 (12.0) 94.6 (5.3) 88.1 (14.6) HS (t2) 97.6 (2.7) 94.3 (9.5) 97.6 (3.5) 89.3 (17.0) EBP (t1) 96.4 (6.1) 95.8 (3.6) 94.3 (7.6) 89.0 (8.6) EBP (t2) 97.6 (3.5) 97.6 (3.1) 94.6 (6.8) 93.7 (4.5) Table options 3.1.2. fMRI data: ROI analyses Region of interest (ROI) analyses revealed significant Group (EBP, HS) by Stroop (congruent, incongruent) interaction within the right and left inferior frontal gyrus (IFG) [F(1,28)=9.08, P=0.005; F(1,28)=5.74, P=0.02, respectively] and within the left hippocampus but at a lenient threshold for this latter result [F(1,28)=3.77, P=0.06]. Specifically, HS exhibited greater activation of right and left IFG for incongruent (I) compared to congruent (C) condition [F(1,28)=8.13, P=0.008; F(1,28)=6.49, P=0.02, respectively], whereas IFG activation was not modulated by the Stroop condition in EBP [F(1,28)=1.98, P=0.16; F(1,28)=0.71, P=0.41., respectively]. Moreover, trend interaction between Group (EBP, HS) and Valence (positive, negative) conditions within right and left hippocampus (HIP) [F(1,28)=3.23, P=0.08; F(1,28)=3.99, P=0.05, respectively] was observed. In these regions, greater activation for negative (N) compared to positive (P) condition was observed in EBP [F(1,28)=6.50, P=0.02; F(1,28)=6.56, P=0.02, for right and left hippocampus, respectively], whereas no significant difference was observed in HS [F(1,28)=0.01, P=0.94; F(1,28)=0.12, P=0.73., respectively] ( Fig. 2). No main Group effect was observed in each ROI. Moreover, no significant difference between EBP and HS was obtained for the right and left amygdala. Results of the ROIs analyses. Panel A shows % MR signal intensity variation in ... Fig. 2. Results of the ROIs analyses. Panel A shows % MR signal intensity variation in the left and right inferior frontal gyrus respectively for the Group-by-Stroop (EBP vs. HS * I vs. C) and Time-by-Stroop (t1 vs. t2 * I vs. C) interactions. Panel B shows % MR signal intensity variation in the left and right hippocampus respectively for the Group-by-Valence (EBP vs. HS * N vs. P) and Time-by-Valence (t1 vs. t2 * N vs. P) interactions. Error bars represent standard error. IFG=Inferior Frontal Gyrus; HIP=Hippocampus; EBP=Euthymic Bipolar Patients; HS=Healthy Subjects; t1=before psychoeducation; t2=after psychoeducation; C=Congruent, I=Incongruent; P=Positive; N=Negative; ⁎P<0.05. Figure options 3.1.3. fMRI data: whole brain analyses Differences between HS and EBP (Group contrast) at t1 were evaluated according to the three main contrast effects: Stroop [Incongruent (I)>Congruent (C)], Valence [Negative (N)>Positive (P)] and Arousal [CP+CN+IP+IN (All)>Neutral]. According to ROI analysis, the interaction Group-by-Stroop (HS>EBP ⁎ I>C) revealed significant differences between EBP and HS within triangular part of the right inferior frontal gyrus (IFG, F3T). Additional regions were identified within the left orbital part of the superior frontal gyrus (SFG, F1O), right insula (IN), right parahippocampal gyrus (PHIP), left fusiform gyrus (FUSI), left middle occipital gyrus (O2), left lingual gyrus (LING) and right cerebellum (P<0.001 uncorrected, k>10) ( Table 3, Fig. 3A). The reverse interaction (EBP>HS ⁎ I>C) did not reveal supra-threshold voxel activation. The interaction Group-by-Valence (EBP>HS ⁎ N>P) revealed differences between EBP in HS in the bilateral hippocampus (HIP) and the parahippocampal gyrus (PHIP) as well as in left middle temporal gyrus (MTG, T2) (P<0.005 uncorrected, k>20 voxels) ( Table 3, Fig. 3B). The reverse interaction (HS>EBP ⁎ N>P) did not reveal supra-threshold voxel activation. Finally, the interaction Group-by-Emotional Arousal [EBP>HS ⁎ All>Neutral] and its reverse [HS>EBP ⁎ All>Neutral] did not reveal supra-threshold activated voxels (see Supplementary Appendix A for an illustration of the HRF of the main results). Table 3. Shows activation peaks provided by the random-effect between-group (EBP vs. HS before psychoeducation) and by the random-effect within-group (EBP at t1 vs. EBP at t2) analyses for both the Stroop contrast (I vs. C) and the Valence contrast (N vs. P), labelled according to Tzourio-Mazoyer et al., 2002. HS=Healthy Subjects, EBP=Euthymic Bipolar Patients, t1=before psychoeducation, t2=after psychoeducation, C=Congruent, I=Incongruent, P=Positive, N=Negative, Hem=Hemisphere, R=Right, L=Left, k=number of voxels/cluster. Lobe Region aal-label H x y z t k Interaction Group-by-Stroop (HS>EBP ⁎ I>C) Frontal Inferior frontal gyrus, triangular part F3T R 48 35 13 4.65 23 Superior frontal gyrus, orbital part F1O L −18 65 1 4.53 12 Insular Insula IN R 30 23 −14 3.90 16 Limbic Parahippocampal gyrus PHIP R 21 −40 −8 5.06 25 Occipital Fusiform gyrus/Cerebellum FUSI R 42 −70 −23 4.92 48 Fusiform gyrus FUSI L −27 −46 −20 4.29 17 Lingual gyrus LING L −12 −91 −14 4.03 18 Lingual gyrus LING L −21 −55 −14 3.98 13 Middle occipital gyrus O2 L −33 −91 1 4.40 31 Interaction group-by-valence (EBP>HS ⁎ N>P) Limbic Hippocampus/parahippocampal gyrus HIP/PHIP R 24 −34 −11 3.92 58 Hippocampus HIP L −24 −28 −11 3.81 32 Temporal Middle temporal gyrus T2 L −63 −46 −8 4.09 23 Interaction time-by-Stroop (t2>t1 ⁎ I>C) Frontal Inferior frontal gyrus, triangular part F3T R 51 35 19 7.72 31 Temporal Middle temporal gyrus T2 L −48 −46 7 7.00 67 Parietal Precuneus PQ R 12 −70 40 6.18 22 Occipital Cuneus/fusiform gyrus Q R 9 −88 7 11.16 180 Fusiform gyrus FUSI L −33 −46 −23 5.20 10 Middle occipital gyrus O2 L −27 −79 10 6.64 65 Interaction Time-by-Valence (t2>t1 * N>P) Limbic Hippocampus/parahippocampal gyrus HIP/PHIP R 21 −31 −11 5.02 22 Temporal Middle temporal gyrus T2 L −66 −46 −8 4.12 15 Table options Results of the whole-brain analyses. Panels A and B show between-groups (EBP vs. ... Fig. 3. Results of the whole-brain analyses. Panels A and B show between-groups (EBP vs. HS) differences for Stroop-related (I>C ⁎ HS>EBP) (P<0.001 uncorrected, k>10) and Valence-related activity (N>P ⁎ EBP>HS) (P<0.005 uncorrected, k>20), respectively. Panels C and D show activations variations in EBP before (t1) and after (t2) psychoeducation for Stroop-related (I>C ⁎ t2>t1) (P<0.001 uncorrected, k>10) and Valence-related (N>P ⁎ t1>t2) (P<0.005 uncorrected, k>20) activity, respectively. Activated regions were projected onto lateral (left, right) and axial (bottom) views of a 3D anatomical template and labelled according to Tzourio-Mazoyer et al. (2002). EBP=Euthymic Bipolar Patients; HS=Healthy Subjects; t1=before psychoeducation; t2=after psychoeducation; C=Congruent, I=Incongruent; P=Positive; N=Negative; LH=left hemisphere, RH=right hemisphere. Figure options 3.2. Results HS at t2 vs. HS at t1 The comparison of cerebral activity in HS between t1 and t2 on the three contrasts (i.e., Stroop, Valence and Emotional Arousal) did not reveal significant suprathreshold activation, indicating no effect of repeated testing. 3.3. Results on psychoeducation: EBP at t2 vs. EBP at t1 3.3.1. Clinical and neuropsychological comparisons First, on a clinical level and after the psychoeducation (t1 vs. t2), the MANOVAs revealed significant clinical improvement in EBP on specific characteristics of BD [Wilk′s lambda=0.02; F(4,9)=12.51; P=0.03] as well as on patients’ affective state at a lenient threshold [Wilk′s lambda=0.09; F(4,9)=4.70; P=0.07]. Specifically, the level of depression (MADRS), trait-anxiety (STAI-B) and affect intensity (AIM) has significantly decreased after psychoeducation. Moreover, at t2, EBP showed better knowledge and representation of their disease (scales developed by FACE-BD) and the circadian rhythm was more stable (CSM). They also use more appropriate coping strategies, i.e., less focused on the problem and they looked for more social supports (WCC problem and social support) ( Table 1). Second, on a neuropsychological level, EBP performances were within the norms defined for each test, both at t1 and t2. This suggests that the neuropsychological scales used did not reveal major cognitive impairment in EBP at two experimental time-conditions (Supplementary Table 1). 3.3.2. Task performance We did not observe significant effect of the time (t2 vs. t1, scan repetition) neither on RT [F(1,26)=1.34, P=0.26] nor on % CR [F(1,26)=3.47, P=0.07] indicating no repeated testing effect. Additionally, we did not observe an interaction between group and time neither on RT [F(1,26)=0.00, P=0.97] nor on % CR [F(1,26)=0.28, P=0.60], signifying that EBP had not improved their behavioral performance after psychoeducation. 3.3.3. fMRI data: ROI analyses Region of interest (ROI) analyses revealed significant Time (t1, t2) by Stroop (congruent, incongruent) interaction within right and left inferior frontal gyrus (IFG) [F(1,12)=15.47, P=0.002; F(1,12)=9.41, P=0.01, respectively]. In both left and right IFG, the activity was not modulated by the Stroop condition before the psychoeducation but was greater for incongruent compared to congruent condition at t2 [F(1,12)=12.08, P=0.004; F(1,28)=24.51, P<0.001., respectively] ( Fig. 2). Moreover, a main Time effect was observed in the right hippocampus (HIP) [F(1,12)=5.85, P=0.03]. The activity of the right HIP was significantly lower after the psychoeducation. No main effect of the Time was observed in other ROI. No significant difference between t1 and t2 was observed either in the left HIP or in right and left amygdala. 3.3.4. fMRI data: Whole brain analyses The same contrasts (Stroop, Valence and Arousal) were used to assess differences in brain activity between t1 and t2 in EBP. First, the interaction Time-by-Stroop (t2>t1 ⁎ I>C) in EBP revealed significant increased activity within the triangular part of the right IFG (F3T), the left middle temporal gyrus (T2), the right precuneus (PQ), the right cuneus (Q), the left fusiform gyrus (FUSI) and the left middle occipital gyrus (O2) after psychoeducation (P<0.001 uncorrected, k>10) ( Table 3, Fig. 3C). The reverse interaction (t1>t2 ⁎ I>C) did not reveal supra-threshold activated voxels. Second, the interaction Time-by-Valence (t2>t1 ⁎ N>P) in EBP revealed significantly decreased activity after psychoeducation within the right parahippocampal gyrus (PHIP) and the left middle temporal gyrus (T2) after the psychoeducation (P<0.005 uncorrected, k>20) ( Table 3, Fig. 3D). The reverse interaction (t2>t1 ⁎ N>P) did not reveal supra-threshold activated voxels. Finally, the interaction Time-by-Emotional Arousal (t1>t2 ⁎All>Neutral) and its reverse (t2>t1 ⁎ All>Neutral) did not reveal supra-threshold activated voxels (see Supplementary Appendix A for an illustration of the HRF of the main results).