© 2014 John Wiley & Sons A/S Published by John Wiley & Sons Ltd.

Bipolar Disorders 2014: 16: 830–845

BIPOLAR DISORDERS

Original Article

Episodic memory impairments in bipolar disorder are associated with functional and structural brain changes Oertel-Kn€ ochel V, Reinke B, Feddern R, Knake A, Kn€ ochel C, Prvulovic D, Pantel J, Linden DEJ. Episodic memory impairments in bipolar disorder are associated with functional and structural brain changes. Bipolar Disord 2014: 16: 830–845. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. Objectives: We combined multimodal functional magnetic resonance imaging (fMRI) and structural magnetic resonance imaging to probe abnormalities in brain circuits underpinning episodic memory performance deficits in patients with bipolar disorder (BD). Methods: We acquired whole-brain fMRI data in 21 patients with BD and a matched group of 20 healthy controls during a non-verbal episodic memory task, using abstract shapes. We also examined density of gray matter, using voxel-based morphometry (VBM), and integrity of connecting fiber tracts, using diffusion tensor imaging (DTI) and tractbased spatial statistics, for areas with significant activation differences. Results: Patients with BD remembered less well than controls which shapes they had seen and had lower activation levels during the encoding stage of the task in the anterior cingulate gyrus, the precuneus/cuneus bilaterally, and the left lingual gyrus, and higher activation levels during the retrieval stage in the left temporo-parietal junction. Patients with BD showed reduced gray matter volumes in the left anterior cingulate, the precuneus/cuneus bilaterally, and the left temporo-parietal region in comparison with controls. DTI revealed increased radial, axial, and mean diffusivity in the left superior longitudinal fascicle in patients with BD compared with controls. Conclusions: Changes in task-related activation in frontal and parietal areas were associated with poorer episodic memory in patients with BD. Compared with data from single imaging modalities, integration of multimodal neuroimaging data enables the building of more complete neuropsychological models of mental disorders.

There is increasing evidence that bipolar disorder (BD) is accompanied by considerable and often progressive cognitive dysfunction (1). Cognitive deficits in patients with BD have been documented in the domains of attention, memory, and executive functions (2). Poor episodic memory performance is not restricted to acute manic or depressive episodes but may persist during the

€ chela, Britta Viola Oertel-Kno a Reinke , Richard Fedderna, Annika € chela, David Knakea, Christian Kno a Prvulovic , Johannes Pantelb and David EJ Lindenc a

Laboratory of Neurophysiology and Neuroimaging, Department of Psychiatry, Psychosomatic Medicine and Psychotherapy, Goethe University, bInstitute of General Practice, Goethe University, Frankfurt, Germany, cMRC Centre for Neuropsychiatric Genetics and Genomics, Institute of Psychological Medicine and Clinical Neurosciences, School of Medicine, Cardiff University, Cardiff, UK doi: 10.1111/bdi.12241 Key words: bipolar disorder – DTI – fMRI – non-verbal episodic memory – VBM Received 16 April 2013, revised and accepted for publication 20 December 2013 Corresponding author: Viola Oertel-Kno€chel Laboratory of Neurophysiology and Neuroimaging Department of Psychiatry, Psychosomatic Medicine and Psychotherapy Goethe University Heinrich-Hoffmann-Str. 10 Frankfurt 60528 Germany Fax: +49-69-6301-3833 E-mail: [email protected]

remitted state (1). Most of the existing research on episodic memory performance in BD has focused on learning- and recall tests in the verbal domain (3). Yet, episodic memories are often stored nonverbally as visual images (4). An exclusive use of verbal memory tasks may thus not capture the full picture of altered episodic memory performance in BD. We therefore aimed to specifically investigate

Correction made after online publication on September 16:Results section has been updated.

830

Episodic memory in bipolar disorders non-verbal episodic memory in remitted patients with BD. Recent findings (2, 5) indeed suggest that remitted patients with BD may perform significantly worse during immediate and delayed nonverbal episodic memory tests in comparison with healthy controls. The structural and functional underpinnings of episodic memory deficits observed in patients with BD are still a matter of debate (6, 7). Across various cognitive tasks, patients with BD show converging alterations of task-related activation in prefrontal and limbic areas (8–10). These findings apply to both working memory processing (6, 11–14) and autobiographical recall (15). Two studies specifically examined the functional activation pattern of episodic memory in BD using positron emission tomography (PET) (16, 17). Both studies showed poor verbal memory performance in patients with BD in relation to prefrontal hypometabolism (dorsolateral prefrontal cortex) (16, 17) and temporal hypermetabolism (hippocampus, parahippocampal gyrus, and superior temporal gyrus) (17). Brooks et al. (17) suggested that prefrontal hypo- and limbic hyper-metabolism might contribute to verbal memory deficits in patients with BD. In addition to the functional imaging research, there is a large body of structural magnetic resonance imaging (MRI) studies in BD. A recent meta-analysis (18) suggested structural deficits which most consistently converge to the prefrontal lobe. Additionally, a systematic review of diffusion tensor imaging (DTI) studies in mood disorders including BD suggested fiber integrity changes mainly in frontal, temporal, and limbic white matter regions (19). A recent meta-analysis of DTI studies in BD (20) revealed three significant clusters of decreased fractional anisotropy (FA) in BD (one right posterior temporo-parietal cluster and two left cingulate clusters). Potential associations between functional and structural alterations can be assessed by performing structural and functional imaging in the same patients. However, one shortcoming of previous work in BD is the lack of integrated imaging approaches. One study (21) examined the functional connectivity pattern in relation to white matter integrity, and showed a significant non-linear relationship between left amygdala functional connectivity and white matter fiber integrity, but this study focused on emotional stimuli. Another multimodal imaging study, in elderly patients with BD (22), reported an association between reduced gray matter volume in anterior limbic areas and reduced fiber integrity in the corpus callosum. Strakowski et al. (23) reviewed exist-

ing functional and structural findings in patients with BD and recommended a consensus model of the illness. In this model, they suggested that disruption in early development (e.g., neuronal pruning and white matter connectivity) within a frontal-limbic network may cause the clinical symptoms of the illness. This approach to BD crucially needs a multimodal neuroimaging program. As yet, the underlying functional and structural alterations of poor memory performance have not been systematically examined. Moreover, it is not known whether any functional or structural markers are related to clinical symptomatology (and thus potentially classified as state markers) or are present in distinct illness states (manic, depressive, remitted/euthymic) and thus reflect traits of the disease. We hypothesized that poor non-verbal episodic memory performance is directly associated with hypoactivity in frontal and hyperactivity in limbic (temporal) brain areas involved in episodic memory processing (16, 17, 24, 25). We further explored whether there is a partial overlap between functional and structural alterations in the respective brain fronto-temporal regions, which may indicate the multimodal nature of brain changes in BD.

Materials and methods Participants

We included 21 euthymic patients with BD (see Table 1 for further details) diagnosed with bipolar disorder I according to DSM-IV criteria (26). All patients were in a remitted state of the illness and were outpatients of the Department of Psychiatry, Goethe-University (Frankfurt, Germany). Patients had been on stable medication for a minimum of four weeks. The psychiatric medication included mood stabilizers (n = 21), and additional antidepressant (n = 9), neuroleptic (n = 12), and anxiolytic (n = 3) agents; no patient fulfilled all necessary criteria for an acute manic or depressive episode or for any comorbid Axis I or II disorder. The control (CON) group (n = 20) was matched with the patient group for age, gender, and educational level. None in the CON group had any positive family history of affective disorder. Exclusion criteria for CON participants were current drug abuse, neurological disease, any history of psychiatric disorders, including Axis I and Axis II disorders according to DSM-IV, and an inability to provide informed consent. All participants also took part in another, larger study assessing verbal episodic memory in a separate experimental session. Participants were pro-

831

Oertel-Kn€ ochel et al. Table 1. Sociodemographic and clinical characteristics and cognitive performance of the patients with bipolar disorder (n = 21) and the control group (n = 20)

Variables Gender, female/male, n Age, years Education, years Parental education, years Female group Mother Father Male group Mother Father Handedness Duration of illness, years No. of depressive episodes No. of manic episodes No. of episodes of illness Duration of medication, years BDI-II score BRMAS score PANAS positive score PANAS negative score SCL-90-R score Global severity index Depression MWT-B score TMT-A score CVLT score Delayed free recall I Delayed free recall II Yes/no retrieval Non-verbal episodic memory task No. of hits (%) d0 Reaction time (msec)

Bipolar disorder Mean (SD)

Controls Mean (SD)

9/12 35.67 (10.68) 14.86 (2.43)

8/12 36.90 (11.06) 15.85 (1.84)

v² = 0.02a; ns t = 0.36b; ns z = 1.47c; ns

13.08 (3.27) 13.81 (4.06)

12.88 (3.41) 14.60 (2.45)

z = 0.43c; ns

13.43 (3.31) 14.24 (4.10) All left-handed 7.62 (5.82) 7.70 (10.97) 6.72 (7.15) 13.75 (12.00) 6.26 (6.09) 9.85 (8.97) 0.38 (0.59) 25.62 (8.34) 15.71 (5.25)

13.35 (3.57) 14.70 (2.58) All right-handed

z = 0.81c; ns

2.00 (3.57) 0.25 (0.44) 35.75 (4.61) 14.20 (4.01)

z z z z

= = = =

3.64c; p < 0.01 0.80c; ns 4.78c; p < 0.01 1.03c; ns

0.61 (0.52) 0.84 (0.75) 29.86 (3.31) 34.19 (13.33)

0.16 (0.14) 0.15 (0.15) 31.80 (2.80) 27.75 (8.23)

z z t t

= = = =

3.73c; p < 0.01 3.99c; p < 0.01 1.97b; p = 0.06 1.85b; p = 0.07

11.29 (2.76) 12.33 (3.02) 15.10 (1.14)

13.55 (1.87) 14.05 (1.70) 15.60 (0.71)

t = 3.06b; p < 0.01 t = 2.23b; p < 0.05 t = 1.72b; p = 0.09

81.49 (9.12) 5.90 (2.28) 1,952.56 (664.76)

85.75 (9.97) 7.55 (1.90) 1,908.39 (648.82)

t = 3.28b; p < 0.05 t = 2.502b; p < 0.05 t = 0.12b; ns

Significance

Values are presented as [mean standard deviation (SD)] except for gender, which is presented as ‘n’. We corrected all statistical data for multiple comparisons using the Bonferroni correction. BDI-II = Beck Depression Inventory–II; BRMAS = Bech–Rafaelsen Mania Scale; CVLT = California Verbal Learning Test; MWT-B = Mehrfachwahl–Wortschatz-Test; ns = not significant; PANAS = Positive and Negative Affect Schedule; SCL-90-R = Symptom Checklist of Derogatis; TMT-A = Trail Making Test–Part A. a Chi-quadratic test. b t-test. c Mann–Whitney U-test.

vided with a description of the study and gave written informed consent before participating. Experimental procedures were approved by the ethics board of the medical department of Goethe University (Frankfurt, Germany). Assessment of performance

psychopathology

and

cognitive

The Structured Clinical Interview for DSM-IV [SCID-I and SCID-II); German version (27)] was carried out with all participants. Current psychopathology was assessed using the German version of the Beck Depression Inventory–II (BDI-II) (28) and the German version of the Bech–Rafaelsen

832

Mania Scale (BRMAS) (29). Remitted state was defined as a patient having a BDI-II score of < 18 and a BRMAS score of < 7. All participants were also screened for their current emotional and mental state using the Positive and Negative Affect Schedule (PANAS) (30) (positive affect and negative affect subscales) and the Symptom Checklist of Derogatis (SCL-90-R) (31) [subscale depression (SCL-90-R DEP) and global severity index]. In order to assess episodic memory performance, we conducted the California Verbal Learning Test (CVLT) (32), using the parameters learning sum [(LS); sum of the free recall over all five runs], Yes/No recognition (JNW), false positive answers (FP), delayed free-recall I (DFR-I) and delayed

Episodic memory in bipolar disorders or not the recognition item was part of the learning list. We conducted two fMRI runs of four minutes each, in which an encoding phase was followed by a retrieval phase. At the beginning (lasting 16 sec) and at the end (lasting 16 sec) of each trial, a fixation cross was presented, and stimuli were then interleaved with this passive fixation cross, with a presentation duration of two seconds and an interstimulus interval that varied between eight and 12 sec. The participants were asked to focus the fixation cross during the whole assessment. During the five-min break between learning and retrieval, another task (with verbal material; results not reported here) was conducted, which did not interfere with the non-verbal task because of the use of different stimulus material. All participants saw the same stimuli. The participants were instructed not to engage in speech during stimulus periods; this was monitored by the investigator through the scanner microphone. The whole session lasted for approximately 20 min. We computed the following scores: accuracy (% sum of correct responses), overall reaction time, and d-prime (d 0 ). d 0 is a statistical score indicating the separation between the means

free-recall II (DFR-II). We also tested all participants on the Mehrfachwahl–Wortschatz-Test (MWT-B) (33) (the German equivalent of the ‘Spot-the-Word test’) and the Trail Making Test–Part A (TMT-A) (34) in order to control for effects of general intelligence and psychomotor speed. Experimental paradigm

We assessed non-verbal episodic memory performance using a computer-based non-verbal learning and recognition test (see Fig. 1), using the software Presentation (http://www.neurobs.com/). The fMRI task had two phases, encoding and retrieval. During the encoding phase, participants saw a sequence of ten abstract geometrical figures and were instructed to memorize them as accurately as possible (see examples in Fig. 1). During this period, the participants were asked to press a button to indicate if an item had been memorized or not. During the retrieval phase, these stimuli were presented again together with the same number of distractor figures. The participants were instructed to indicate by button press with their right index finger, as accurately and fast as possible, whether

A

ISI: 8–12 sec

2 sec

16 sec BD B

88 87 86 85 84 83 82 81 80 79 78 77

1980

CON

1960 1940 1920 1900 1880 1860

Fig. 1. (A) Example of the experimental paradigm of the non-verbal episodic memory test. The conditions of learning, recognition, and fixation cross alternated with each other. (B) Task performance for verbal episodic memory across groups (t-test for group comparison, scores in %). Controls (CON): n = 20, white bars. Patients with bipolar disorder (BD): n = 21, black bars. ISI = inter-stimulus interval.

833

Oertel-Kn€ ochel et al. of a signal and the noise distributions. It is calculated using the following formula: d 0 = Z (hit rate)  Z (false alarm rate) (35). A higher d 0 indicates that the signal can be more readily detected. Post-scanning debriefing

The results of the post-scanning debriefing were computed by assessing group differences using non-parametric statistical tests. No significant group differences were found on the level of their personal mental state or their attention (p > 0.05). Furthermore, we asked for strategies to recall the intended items. The items of the post-scanning debriefing included questions asking if the objects were associated with any further meaning and questions about possible grouping of items in the encoding period. These items showed no group differences (p > 0.05). Functional data acquisition

Within a week of the first session (which comprised the baseline cognitive assessments), all subjects underwent functional and anatomical imaging on a Siemens Magnetom Allegra 3 Tesla MRI system (Siemens Medical Systems, Erlangen, Germany) at the Frankfurt University Brain Imaging Center (Frankfurt/Main, Germany). Each session included four functional scans [echo planar imaging (EPI) sequence; 465 volumes, voxel size = 3 9 3 9 3 mm; repetition time (TR) = 2,000 msec; echo time (TE) = 30 msec; 33 slices, slice thickness = 3 mm; distance factor = 20%; flip angle = 90°], each lasting approximately four minutes. During the functional scans, participants were scanned with eyes open in darkness and were instructed to look at a fixation cross at the middle of a screen. A mirror was fixed on the head coil in order to view the screen on which the stimulus material was presented in the scanner. We synchronized the stimulus presentation with the fMRI sequence at the beginning of each trial using MRI trigger pulses that triggered the ongoing presentation of stimuli in the Presentationâ software. After the imaging, we conducted a postscanning debriefing, asking the participants to rate their attention, concentration, and mental state during the scan on a five-point Likert scale. Between the first and the second part of the non-verbal trial, we performed an anatomical measurement [modified driven equilibrium fourier transform (36); voxel size = 1 9 1 9 1 mm, 176 slices]. Moreover, a diffusion tensor imaging session was conducted using an EPI sequence [TR = 8,760 msec; TE = 100 msec; bandwidth

834

1,302 Hz/pixel; acquisition voxel size = 2 9 2 9 2 mm3; 60 axial adjacent slices, slice thickness = 2 mm (no gap); field of view (FOV) = 192 mm 9 192 mm 9 120 mm; acquisition matrix = 96 9 96; ten images without diffusion weighting (b0) with 60 diffusion-encoded images (b-values = 1,000 sec/mm2 60 non-colinear directions)]. The diffusion MRI images were averaged (total acquisition time = 10 min). This sequence used parallel acquisition of independently constructed images using generalized auto-calibrating parallel acquisitions (GRAPPA) (37). MRI procedures: functional imaging preprocessing

We pre-processed and analyzed the fMRI data using BrainVoyager QX software, version 2.3 (Brain Innovation, Maastricht, the Netherlands). For the functional data, we applied the following pre-processing steps: slice-time correction, motion correction, linear trend removal, and highpass temporal filtering of two cycles per time course. Co-registration of the functional data to the anatomical scans was performed using automated scripts in BrainVoyager QX, following by manual control of the anatomical alignment. We transformed the three-dimensional (3D) anatomical scans into Talairach space (38) using a 12-point affine transformation as implemented in the BrainVoyager software, and subsequently used the parameters of this transformation to transform the co-registered functional data. We then resampled the 3D functional data set to a voxel size of 1 9 1 9 1 mm3. The protocol of the task was convolved with a hemodynamic response function (39) to generate the predictors for the general linear model (GLM) separately for the encoding and retrieval conditions. For the retrieval predictor, we only included correct trials, and for the encoding predictor only items that were correctly recognized at retrieval. MRI procedures: region-of-interest (ROI) analysis with voxel-based morphometry (VBM)

The VBM pre-processing and statistical analysis were performed using statistical parametric mapping (SPM) 8 (Wellcome Department of Imaging Neuroscience, London, UK) running on MATLAB version 7.7.0. We used the default parameters for pre-processing steps as provided by the SPM software. The chosen smoothing parameters reflected the standard of group studies, whereby a certain amount of smoothing is required to take account of inter-individual anatomical differences (40). First, all images were checked for

Episodic memory in bipolar disorders artifacts, structural abnormalities, and pathologies. Second, customized T1 templates and prior images of gray matter, white matter, and cerebrospinal fluid were created from all participants in order to use them for the group analysis. We used modulated data and prior probability maps (voxel intensity) to guide segmentation in SPM. The segmentation included six different tissue types, light bias regularization (0.001), bias full width at half maximum (FWHM) cut-off 60 mm, warping regularization of 4, affine regularization to the European brain template (linear registration) and a sampling distance of 3 mm. The segmentation was checked for quality before further analysis. Finally, the images were smoothed (40) with a Gaussian kernel of 8 9 8 9 8 mm3 (FWHM), whereby the intensity of each voxel was replaced by the weighted average of the surrounding voxels, in essence blurring the segmented image. We then created ROI masks from the functional imaging task. We extracted the Talairach coordinates (peak voxel) of all significant regions in the group contrast during the non-verbal episodic memory paradigm and created masks not extending 1,000 mm3 (default) using the WFU PickAtlas toolbox (41). For mask creation, Talairach coordinates of ROIs were transformed to Montreal Neurological Institute (MNI) coordinates using the WFU PickAtlas toolbox (41). Here, the selected regions were the anterior cingulate bilaterally, the precuneus/cuneus bilaterally, the left lingual gyrus, and the left middle and superior temporal/inferior parietal gyrus. We compared the differences in volume between individual images. Differences between the groups in the ROIs of gray matter using VBM were then tested using linear statistical contrasts, resulting in a t statistic for each voxel. The respective global volumes of gray and white matter and cerebrospinal fluid as obtained during segmentation were included as nuisance variables. DTI procedures: ROI analysis using the Oxford Centre for Functional MRI of the Brain (FMRIB) Software Library (FSL)

Diffusion MRI data were pre-processed and analyzed using the standard procedure of the TBSS software using FSL 4.1 (http://www.fmrib.ox.ac. uk/fsl) (42). TBSS is a specific voxel-wise approach to analyzing diffusion MRI data, which projects the individual DTI parameters of each participant onto a mean white matter skeleton mask (42). Pre-processing of the diffusion MRI data included motion correction and correction for eddy current distortion. For the motion correction step, all DTI images were quality checked

(43) by two investigators (BR and VO-K) and corrupted volumes (spikes, motion) were removed. The eddy current distortion was done using an inhouse automatic routine, after which all files were averaged into one single 3D data set for each subject. This was followed by warping procedures and non-linear registration for all images, using the TBSS routine (42). The maps for all subjects were transformed to MNI space by non-linear alignment to the FMRIB58_FA, a standard template provided by FSL. A mean FA image was then created and thinned to obtain a mean FA skeleton representing the centers of all tracts common to the group, and all corrected images were fitted using a tensor model that generated the diffusion maps [FA, mean diffusivity (MD), radial diffusivity (DR), axial diffusivity (DA)] used in the subsequent TBSS analysis. Each subject’s aligned FA (and consequently also MD, DR, and DA) map was then projected onto this skeleton (by using an FA skeletonization program) and fed into voxelwise cross-subject statistics using permutation testing [randomize tool in FSL, 5,000 permutations (44)]. This was followed by a similar procedure for examining MD, DR, and DA scores for all participants. After pre-processing of the DTI parameters, the resulting FA, MD, DR, and DA were used for the ROI analyses of all tracts connecting to the gray matter regions that were significantly altered in the group contrast of the fMRI paradigm (bilateral: cingulum, uncinate fascicle, inferior fronto-occipital fascicle, inferior longitudinal fascicle, superior longitudinal fascicle). To prepare the ROI analysis, we extracted tracts around the significant clusters of the FA group analysis, using the JHU whitematter tractography atlas provided by FSL (45). For two ROIs, the truncus and splenium of the corpus callosum, there were no atlas-based masks available, and therefore they were manually generated by two independent raters. For all ROIs, we extracted FA, MD, DR, and DA values for each participant. Statistical analysis

Group comparisons. We computed t-tests assessing group differences in the scores of the CVLT, the MWT-B, the TMT-A, and the non-verbal episodic memory scores (accuracy, reaction time, d 0 ) using SPSS 19.0 software (Statistical Package for Social Sciences, www.spss.com). We also computed statistical tests of variance, including BDI-II, BRMAS, PANAS, and SCL90-R scores as dependent variables and the participant group as the fixed factor. Because these

835

Oertel-Kn€ ochel et al. data were not normally distributed, we selected a non-parametric statistical test (Mann–Whitney U-test). We also performed an additional analysis of covariance, using the scores for symptoms of depression (BDI-II) and mania (BRMAS) as covariates and the non-verbal episodic memory scores as the dependent variable. An F-statistic was calculated for each voxel to create whole-brain activation maps using BrainVoyager QX 2.3 software. For this model, we included group as the fixed factor, with two levels (CON group, patients with BD), and activation maps as dependent variables. Group activation maps [thresholded at q < 0.01, false discovery rate (FDR)-corrected for multiple comparisons (46) unless stated otherwise] were created to identify regions significantly activated during the episodic memory conditions compared to baseline. We then computed first- and second-level comparisons. On the first level, we performed contrast analyses between the encoding and retrieval conditions and their respective ‘fixation cross’ baseline conditions. On the second level, we compared the activation maps of encoding and retrieval across groups (CONs versus patients). The threshold for the firstlevel analysis was fixed at an alpha level of 5%, corrected for multiple comparisons using the FDR correction. The threshold for group contrasts was fixed at an alpha level of 5% using the cluster-level thresholding tool implemented in BrainVoyager QX, with 1,000 iterations. The resulting statistical maps of the ROI gray matter analysis (VBM) using the SPM8 package showed all voxels of the ROIs with showed a significant group difference between all voxels of the ROIs (minimum cluster size = 100 mm3; p-value thresholded at p < 0.001, small volume correction). The significant results of the analysis were interpreted as volumes differences between the groups. We also computed t-tests assessing group differences in ROI FA, MD, DR, and DA values from the DTI analysis using the FSL software (at an alpha set of p < 0.05, corrected for multiple comparisons using the Bonferroni correction). The individual scores were extracted in FSL and entered into an SPSS data file, which was used for further analyses. Correlation analyses. We conducted correlation analyses in order to test whether the main functional and structural results were directly associated with clinical or cognitive symptoms, using SPSS 19.0 software. We used non-parametric correlations (Spearman rank correlation) if the data were not normally distributed, and parametric

836

correlations (Pearson product-moment correlation) if the data were normally distributed. We corrected all correlations for multiple comparisons using Bonferroni correction with a corrected alpha level < 0.05). All correlation analyses were computed for each subject group independently, to avoid confounding the effects of group differences on correlation strengths. We performed bivariate correlation analyses between clinical (BDI-II, BRMAS) and cognitive (non-verbal episodic memory performance, CVLT) scores and beta values (fMRI), volumes, and DTI scores. For this analysis, we selected only the areas which showed significant group differences during group comparisons. In order to rule out potential partial volume effects, we computed bivariate correlation analysis between the beta scores of significantly activated areas during memory-related fMRI, volumes, and DTI scores. We performed bivariate correlation analyses between the beta values of fMRI, gray matter volumes, and DTI scores and the medication doses were computed according to the method of Almeida et al. (47). We also provided an additional correlation analysis using the years of medication and the beta values of the functional activation during episodic memory retrieval.

Results Individual psychopathology and cognitive performance

The statistical tests (Mann–Whitney U-test) for group differences showed significantly higher levels of depressive symptoms (BDI-II) in the BD group in comparison with the CON group [BD: mean = 9.85, standard deviation (SD) = 8.97; CON: mean = 2.00 (SD = 3.57); z = 3.64, p < 0.01], but no group differences in the scores of the manic scale (BRMAS) [BD: mean = 0.38 (SD = 0.59); CON: mean = 0.25 (SD = 0.44); z = 0.80, p > 0.05]. Further tests for group differences showed significant group differences for the PANAS positive affect [BD: mean = 25.62 (SD = 8.34); CON: mean = 35.75 (SD = 4.61); z = 4.78, p < 0.01], but no group differences in the PANAS negative affect [BD: mean = 15.71 (SD = 5.25); CON: mean = 14.20 (SD = 4.01); z = 1.03, p > 0.05]. The SCL-90-R depression subscale and global severity index scores were significantly higher for the patient group: (i) depression: [BD: mean = 0.84 (SD = 0.75); CON: mean = 0.15 (SD = 0.15); z = 3.99, p < 0.01] and (ii) global severity [BD: mean = 0.61 (SD = 0.52); CON: mean = 0.16 (SD = 0.14); z = 3.73, p < 0.01] (see Table 1).

Episodic memory in bipolar disorders There was no significant difference in the t-test assessing group differences in the MWT-B [BD: mean = 29.86 (SD = 3.31); CON: mean = 31.80 (SD = 2.80); t = 1.97, p > 0.05]. The test for group differences in the TMT-A revealed slower psychomotor speed for patients with BD in comparison with CONs, but the difference showed only trend-level significance [BD: mean = 34.19 (SD = 13.33); CON: mean = 27.75 (SD = 8.23); t = 1.85, p = 0.07]. CONs also performed significantly better than patients with BD in the subscale DFR-I and DFR-II of the CVLT (DFR-I: t = 3.06, p < 0.01; DFR-II: t = 2.23, p < 0.01) (for details, see Table 1). In comparison with patients with BD, CONs showed significantly higher accuracy in the non-verbal episodic memory task, regarding the number of hits [BD: mean = 81.49 (SD = 9.12); CON: mean = 85.75 (SD = 9.97); t = 3.28, p < 0.01] and the d 0 score [BD: mean = 5.90 (SD = 2.28); CON: mean = 7.55 (SD = 1.90); t = 2.502, p < 0.01]. There was no group difference in the reaction time during the episodic memory paradigm (p > 0.05). fMRI results

Effects for encoding and retrieval. We compared activation during encoding and retrieval to the in order to ascertain whether the canonical memory networks were activated by the task. We superimposed these activation maps on a surface-reconstructed MNI template in BrainVoyager QX. For encoding, significant activation in contrast to baseline across the groups was found in bilateral frontal (medial, superior, inferior), bilateral parietal and occipital (precuneus, cuneus, lingual gyrus, occipital gyrus), and bilateral temporal (fusiform gyrus, parahippocampal gyrus) regions. Significant deactivation during encoding across the groups was observed bilaterally in the anterior medial frontal gyrus, the left posterior insula, and the left anterior and bilateral posterior cingulate gyrus [thresholded at q < 0.01, FDR-corrected for multiple comparisons (46)] (see Supplementary Fig. 1). During retrieval, the bilateral frontal (medial, middle, superior, inferior gyrus, insula), bilateral precuneus, and inferior parietal lobe; bilateral occipital (cuneus, lingual gyrus, inferior and middle occipital lobe) and bilateral temporal (fusiform, parahippocampal, superior temporal gyrus) regions; and bilateral anterior and posterior cingulate gyrus were significantly active in contrast to baseline across groups. Significant deactivation during retrieval in comparison with baseline across the groups was found in the right anterior

medial frontal gyrus and bilaterally in the inferior anterior cingulate gyrus and right inferior parietal lobe [thresholded at q < 0.01, FDR-corrected for multiple comparisons (46)] (see Supplementary Fig. 1). Group comparisons for encoding and retrieval. The group contrast (t-test) for encoding (thresholded at cluster level a ≤ 0.05%, 1,000 iterations, 10,831 mm3) showed higher activation in CONs in comparison with patients in the anterior cingulate bilaterally (left: t = 2.97, p < 0.01; right: t = 4.48, p < 0.01), in the precuneus/cuneus bilaterally (left: t = 4.83, p < 0.01; right: t = 7.22, p < 0.01), and the left lingual gyrus (t = 4.93, p < 0.01) (see Fig. 2 and Table 2). The group comparison (contrast t-maps) for episodic memory retrieval (thresholded at cluster level a ≤ 0.05%) showed higher deactivation in CONs in comparison with patients in a left temporo-parietal region, including the middle and superior temporal gyrus and the inferior parietal lobe (angular and supramarginal gyri) (t = 3.508, p < 0.01) (Fig. 2, Table 2). These group differences remained significant after controlling for individual performance levels on the episodic memory task. Post-hoc ROI analysis: hippocampus-related group contrasts. We added a post-hoc ROI analysis with a bilateral hippocampus mask provided by BrainVoyager QX (48,631 lL; x = 20, y = 30, z = 4) because of the general importance of this area for episodic memory processes. We computed activation contrasts for the encoding and retrieval conditions. We found significant activation within the hippocampus bilaterally during encoding and retrieval. The pattern of results indicated that the ventral part of the hippocampus bilaterally showed increased activity during encoding [left: t(39) = 4.510, p < 0.01; right: t(39) = 5.761, p < 0.01] and retrieval [left: t(39) = 5.021, p < 0.01; right: t(39) = 4.808, p < 0.01]. The dorsolateral part of the hippocampus bilaterally showed deactivation during both task phases: (i) encoding [left: t(39) = 1.621, p < 0.01; right: t(39) = 1.511, p < 0.01] and (ii) retrieval [left: t(39) = 5.021, p < 0.01; right: t(39) = 4.808, p < 0.01] [all p-values thresholded at q (FDR) < 0.01 (46)]. Post-hoc group contrasts showed higher activation in CONs in comparison with patients with BD within the ventral hippocampus bilaterally, mostly located in the dorsomedial (encoding, retrieval) and ventrolateral (retrieval) parts of the hippocampus (thresholded at cluster level a ≤ 0.05%) (see Fig. 3 and Supplementary Table 1).

837

Oertel-Kn€ ochel et al. Gray matter: ROI analysis with VBM

Statistical tests for group differences [t-tests, cluster size: 100 mm³, p < 0.05 (FDR-corrected)] between patients with BD and CONs in the ROIs (anterior cingulate bilaterally, precuneus/cuneus bilaterally, left lingual gyrus, left middle and superior temporal gyrus) revealed significantly lower amounts of gray matter in patients with BD compared with CONs in the left anterior cingulate (t = 4.09, p < 0.01), the precuneus/cuneus bilaterally (left: t = 4.94, p < 0.05; right = 5.11, p < 0.01), and the temporo-parietal junction (t = 5.45, p < 0.01) (see Table 2, Fig. 2). Gray matter volumes in the right anterior cingulate (t = 1.23, p > 0.05) and the left lingual gyrus (t = 1.03, p > 0.05) were not significantly different across groups. DTI ROI analysis

ROI analysis of tracts connecting the areas which showed significant group differences in fMRI data included the bilateral cingulum, uncinate fascicle, inferior fronto-occipital fascicle, and inferior and superior longitudinal fascicle. Patients with BD showed significantly higher MD, DR, and DA scores in comparison with CONs in the left superior longitudinal fascicle (MD: t = 8.66, p < 0.01; DR: 8.44, p < 0.01; DA: 6.39, p < 0.01) (Table 2, Fig. 2). FA scores were not significantly different between groups (left: t = 1.07, p > 0.05; right: t = 0.50, p > 0.05). There were no significant group differences in the DTI scores of the right superior longitudinal fascicle, the bilateral cingulum, the uncinate fascicle, the inferior fronto-occipital fascicle, or any of the other investigated tracts (p > 0.05 for all). Correlation analysis

The beta scores of the lingual gyrus (extracted from the group contrast of the fMRI task) were significantly correlated with the FA scores of the inferior fronto-occipital fascicle in the patient group (r = 0.50, p < 0.05). However, none of the other correlation analyses between DTI, fMRI,

and VBM parameters showed any significant associations. Functional activation levels in the left lingual gyrus were significantly correlated with the ‘yes/no retrieval’ score of the CVLT both in CONs and patients with BD (CON: r = 0.49, p = 0.009; BD: r = 0.56, p = 0.008) and with non-verbal episodic memory performance in patients only (BD: r = 0.61, p = 0.005; CON: r = 0.20, p > 0.05). We also conducted a Fisher’s Z for difference between correlation coefficients and showed that the correlations coefficients between groups were significantly different (Z = 2.012, p = 0.04). None of the assessed DTI or VBM scores showed significant correlations with the non-verbal episodic memory performance (p > 0.05). We further tested whether clinical symptom severity and medication status may have been significantly associated with the fMRI, VBM, or DTI results. None of these correlations reached significance (p > 0.05). Discussion

Remitted patients with BD, compared with healthy CONs, showed lower non-verbal episodic memory performance. This is in line with previous findings of significantly worse performance of non-acute patients with BD in non-verbal episodic memory tests in comparison with healthy CONs (2, 5). A possible explanation for this finding is that patients still showed residual psychopathology, albeit not reaching cut-off values for depressive (BDI-II) or manic (BRMAS) episodes. This residual psychopathology might have affected cognitive performance in patients with BD. However, additional analysis of covariance with the depressive or manic test values revealed no significant influence of these parameters on the current test results. The brain activation pattern during memory encoding and retrieval revealed in our study conforms broadly to the task-related activation patterns observed in previous functional imaging studies (16, 17, 24, 25). Crucially, our comparisons of brain activation during encoding and retrieval

Fig. 2. t-tests of the group contrast controls (n = 20) versus BP patients (n = 21) (t-map cluster-level corrected, q < 0.05) during fMRT-task sequences. Upper row (A): Regions with significant activation differences between controls and BP patients during encoding. Second row (B): Regions with significant activation differences between controls and BP patients during retrieval. Color code: red indicates con > pat, blue indicates pat > con. TG = temporal gyrus. The left side in the figure indicates the right side of the brain (radiological convention). (c) Grey matter volume differences in ROIs of BD patients (n = 21) versus controls (n = 20) (VBM) (CON > PAT marked in yellow). Minimum cluster size was > 100 voxels. The images are in neurological convention (right = right). The color codes denote t-scores of all relevant areas. The non-significant ROIs are not shown in the figure. (d) DTI ROI analysis: The left superior longitudinal fascicle showed differences in BD patients versus controls (for details see text and Table 2). Color code: green: White matter skeleton mask. Red: significant differences in this fiber tract during group contrasts.

838

Episodic memory in bipolar disorders ant. cingulate

A

CON > PAT: red PAT > CON: blue 8

cuneus

Ling. gyrus

precuneus

0

B

middle TG z=0 C

z = 10

ROI mask

STG z = 20

CON > PAT

Inf. parietal lobe z = 40

ROI mask

z = 50

–8 t(39) p < 0.05

CON > PAT (marked yellow)

l. ant. cingulate

l. precuneus

l.temporopolar

r. precuneus

D

green: white matter skeleton mask red: significant fiber trakt during group contrasts

l.sup. long-itudinal fascicle

839

Oertel-Kn€ ochel et al. Table 2. Group comparison (t-tests) between controls (CON) (n = 20) and patients with bipolar disorder (BD) (n = 21) during non-verbal episodic memory encoding and retrieval (p < 0.05, cluster-level corrected)a Talairach coordinates Area fMRI Non-verbal encoding Anterior cingulate

Precuneus Cuneus

Lingual gyrus Non-verbal retrieval Middle and superior temporal gyrus Inferior parietal lobe

BA

y

L

1

28

8

260

R

8

35

2

839

L

5

73

28

3,859

R

7

67

32

4,496

18, 19

L

30

72

1

963

39, 40

L

44

56

24

2,276

7, 31 18, 19

z

b scores Mean (SD)

Cluster size (voxel)

x

24, 32

Hemisphere

t-test

CON > BD

CON: 0.13 (0.55) BD: 0.55 (0.87) CON: 0.22 (0.78) BD: 1.00 (0.95) CON: 1.64 (1.09) BD: 0.02 (1.12) CON: 1.34 (0.60) BD: 0.58 (1.05) CON: 2.28 (1.25) BD: 0.55 (0.98)

2.97

p < 0.01

CON: 1.30 (1.29) BD: 0.14 (1.33)

3.51

BD > CON

4.48 4.83

p < 0.01

7.22 4.93

p < 0.01 p < 0.01

MNI coordinates Area Gray matter Non-verbal encoding Anterior cingulate

Precuneus Cuneus

Lingual gyrus Non-verbal retrieval Middle and superior temporal gyrus Inferior parietal lobe

BA

Hemisphere

y

L

6

34

0

213

R

8

33

0

205

L

5

76

26

432

R

7

67

32

541

18, 19

L

30

72

1

963

39, 40

L

44

58

23

569

24, 32

7, 31 18, 19

z

b scores Mean (SD)

Cluster size (voxel)

x

t-test

CON > BD

4.09

p < 0.01

1.23

ns

4.94

p < 0.05

5.11

p < 0.05

1.03

ns

5.45

p < 0.01

DTI scores

t-test

CON > BD

CON: 034 (0.01) BD: 0.33 (0.02) CON: 0.36 (0.02) BD: 0.35 (0.02) CON: 0.76 (0.02) BD: 0.78 (0.03) CON: 0.76 (0.03) BD: 0.76 (0.04) CON: 0.63 (0.02) BD: 0.65 (0.03) CON: 0.62 (0.03) BD: 0.62 (0.04) CON: 1.02 (0.02) BD: 1.03 (0.03) CON: 1.03 (0.03) BD: 1.03 (0.04)

1.07

ns

0.50

ns

8.66

p < 0.01

0.15

ns

8.44

p < 0.01

0.26

ns

6.38

p < 0.05

0.00

ns

CON: 0.55 (0.21) BD: 0.38 (0.18) CON: 0.50 (0.11) BD: 0.46 (0.30) CON: 0.65 (0.22) BD: 0.52 (0.09) CON: 0.60 (0.23) BD: 0.49 (0.20) CON: 0.45 (0.13) BD: 0.40 (0.15) CON: 0.45 (0.17) BD: 0.35 (0.14)

BD > CON

MNI coordinates ROI

Scores

Hemisphere

ROI–DTI analysis Superior longitudinal fasciculus

FA

MD

y

z

L

41

25

26

13,814

R

40

22

30

11,098

L R

DR

L R

DA

Cluster size (voxels)

x

L R

For voxel-based morphometry and DTI–ROI analyses, we only report those comparisons which showed significant group differences. We transformed the Talairach coordinates from the functional magnetic imaging (fMRI) data into Montreal Neurological Institute (MNI) coordinates using the WFU Pickatlas toolboxâ. Cluster size > 100 mm3. BA = Brodmann area; BD = bipolar disease; CON = control group; DA = axial diffusivity (mm²/sec 9 103); DR = radial diffusivity (mm²/ sec 9 103); DTI = diffusion tensor imaging; FA = fractional anisotropy; L = left; MD = mean diffusivity (mm²/sec 9 103); ns = not significant; R = right; ROI = regions of interest; SD = standard deviation. a In gray matter volumes in ROIs selected from the fMRI activation pattern [p < 0.05 (false discovery rate)] and in DTI parameters (FA, MD, DR, DA) in ROI selected from the fMRI activation pattern [p < 0.05 (Bonferroni corrected)].

840

Episodic memory in bipolar disorders

Encoding

A

8

.

ROI mask z=0

Main effect z=0

CON > PAT z=0

0

Retrieval B

–8

t(39) p < 0.05

ROI mask z=0

Main effect z=0

CON > PAT z=0

Fig. 3. Post-hoc ROI analysis with a bilateral hippocampus mask provided by BrainVoyager QX (48631 voxels; centres of mass: LH: x = 20, y = 30, z = 4, : x = 22, y = 29, z = 2). (A) The upper line shows the condition encoding versus baseline, the lower line shows the condition retrieval versus baseline. The left row of the image shows the bilateral hippocampus mask, the middle row of the image shows the main effect across group for encoding versus baseline within the hippocampus mask [thresholded at p(FDR) < 0.05], the right row of the image shows the contrast map CON vs. PAT (thresholded at p < 0.05, clusterlevel corrected) (red = CON > PAT).

of non-verbal episodic items between patients with BD and CONs revealed a disruption of mainly left fronto-temporal-parietal brain regions (anterior cingulate, precuneus, left lingual gyrus, left temporo-parietal junction), and the left lingual gyrus hypoactivity was directly related to non-verbal episodic memory deficits in patients with BD. Surprisingly, the patients with BD showed reduced limbic activation compared with CONs during encoding of non-verbal episodic memory material. This finding is in contrast to Chen et al.’s (48) meta-analysis

of fMRI studies showing mainly hyperactivity of limbic regions in patients with BD. However, Chen et al. (48) suggested that the limbic hyperactivity in these studies is mainly associated with emotional material and emotion-related paradigms. In accordance with this hypothesis, the meta-analysis by Houenou et al. (49) showed that only fMRI studies with emotional material showed a limbic hyperactivity in patients with BD. Furthermore, beside the role of the anterior cingulate as the brain region involved in emotional processing (50), it has

841

Oertel-Kn€ ochel et al. been acknowledged to be a core structure of episodic memory processing (51). A study examining the underlying functional pattern of working memory performance in BD (13) also showed hypoactivity of limbic/paralimbic regions (parahippocampal gyrus). Therefore, we suggest that the hypoactivity of limbic regions during the encoding of non-verbal episodic memory material in the present study represents a dysfunction of core memory structures directly associated with the learning material. Our multimodal approach allowed us to complement these functional changes with information about structural deficits, which were evident in patients, in the left anterior cingulate and left temporo-parietal junction and in one of the key fiber tracts connecting frontal and temporal areas. Previous investigations showed that microstructural changes in the white matter of frontalsubcortical circuits lead to a disconnection syndrome between frontal and subcortical regions. In their recent review, Schneider et al. (52) reported that white matter alterations in patients with BD, for instance in the corpus callosum, cingulum, prefrontal regions, fornix, and superior longitudinal fasciculus, are related to emotional dysregulation. Moreover, white matter abnormalities of frontal-subcortical circuits in affective disorders might be related to a dysfunction in the networks linking frontal and subcortical regions (20, 53, 54). These network alterations have been associated with clinical symptoms in BD (52). Furthermore, our structural imaging findings of reduced volume in the left anterior cingulate and left temporo-parietal junction confirm those of recent meta-analyses of morphological changes in BD (19). The higher activation in an area with reduced cortical volume, the left temporo-parietal junction, may point to compensatory hyperactivation in order to preserve performance in light of reduced cognitive capacity. Such models of compensatory hyperactivation have been introduced for dementia (55) and schizophrenia (56) but not previously for BD. Patients with BD in our sample had higher subclinical depression scores (without fulfilling all the criteria of an acute depressive episode) than CONs, although none of the fMRI, ROI-VBM, or ROI-DTI findings were significantly associated with individual psychopathology ratings. By contrast, some of the episodic memory scores [nonverbal episodic memory, verbal episodic memory (CVLT)] were significantly associated with the imaging findings, mainly in the patient group. To conclude, the fact that, in the present study, all three imaging modalities pointed to alterations

842

in a left-sided fronto-temporal circuit, together with cognitive deficits in non-acute patients with BD, leads to the assumption that non-verbal episodic memory deficits and their underlying neuronal network may be a trait marker of the illness (1, 2). In keeping with this interpretation, we did not find any correlations between imaging parameters and clinical symptom scores. However, the study was neither powered nor specifically designed to assess such correlations because our euthymic patient group showed relatively little variation in their symptom scores. There are several methodological difficulties inherent in studies of medicated patients suffering from relapsing–remitting mental disorders, which we addressed as follows. Some authors suggested that the number of previous affective episodes and of years of illness may be associated with episodic memory performance [e.g. (57)], whereas others did not find any association between memory performance and the number of episodes of illness (58). The present study included an assessment of both the number of episodes and of years of illness and their potential association with the memory performance and imaging results, which showed no significant relationship with these variables. We examined only euthymic patients with BD. This procedure ensured that the current findings were not dependent on different illness states. Furthermore, in order to prevent potential confounding effects on the morphological findings from pharmacological medication, we tested only patients who had been on a stable dosage for at least four weeks prior to testing. We also computed the medication doses according to the method of Almeida et al. (47) and performed a correlation analysis to exclude potential associations between medication doses and cognitive performance. None of the patients was taking benzodiazepines or tricyclic antidepressants at the time of testing. Conclusions

A distinct and novel feature of the current study was the combination of functional and structural imaging markers for the assessment of the neural basis of memory deficits in BD. We demonstrated a spatial overlap of functional and structural alterations in frontal and temporal areas accompanied by signs of reduced microstructural integrity of white matter tracts connecting these areas – primarily the superior longitudinal fascicle. In the present study, the analysis of gray matter density and microstructural connectivity was focused on the areas of ROIs that came out of the functional imaging analysis. Further studies might integrate

Episodic memory in bipolar disorders results from the different imaging modalities into one multivariate pathophysiological model of the disease. Furthermore, our findings fit into the model introduced by Strakowski et al. (23), who associated early disturbances in white matter connectivity and neuronal pruning with the clinical outcome of BD. The present results are important for biological models of bipolar disorder because they allow for comparison with other disorders such as schizophrenia, where alterations in this pathway have also been shown (19), and may thus aid a future biological classification of mental disorders. A better understanding of the underlying biological causes of resistant cognitive symptoms may help in the development of specific therapeutic options – for example, the ‘functional remediation’ introduced by Martinez-Aran et al. (59). Their training is a combined neurocognitive and psychoeducative therapeutic program that aims to treat specific cognitive dysfunctions. Another imagingguided therapeutic development might be fMRIbased neurofeedback, which non-invasively targets the brain regions showing patterns of activation that are directly related to cognitive or clinical symptoms (60, 61).

7.

8.

9.

10.

11.

12.

13.

14.

Acknowledgements MRI was performed at the Frankfurt Brain Imaging Centre, supported by the German Research Council (DFG) and the German Ministry for Education and Research (BMBF; Brain Imaging Center Frankfurt/Main, DLR 01GO0203). VO-K was supported by the Adolf Messer Prize from the Freunde der Universit€at, Frankfurt, Germany.

Disclosures The authors of this paper do not have any commercial associations that might pose a conflict of interest in connection with this manuscript.

References 1. Torres IJ, Boudreau VG, Yatham LN. Neuropsychological functioning in euthymic bipolar disorder: a metaanalysis. Acta Psychiatr Scand 2007; 434: 17–26. 2. Bora E, Yucel M, Pantelis C. Cognitive endophenotypes of bipolar disorder: a meta-analysis of neuropsychological deficits in euthymic patients and their first-degree relatives. J Affect Disord 2009; 113: 1–20. 3. Helmstaedter C, Lendt M, Lux S. Verbaler Lern-und Merkf€ahigkeitstest. G€ ottingen: Hogrefe, 2001. 4. Conway MA. Episodic memories. Neuropsychologia 2009; 47: 2305–2313. 5. Kurtz MM, Gerraty RT. A meta-analytic investigation of neurocognitive deficits in bipolar illness: profile and effects of clinical state. Neuropsychology 2009; 23: 551–562. 6. Townsend J, Bookheimer SY, Foland-Ross LC et al. fMRI abnormalities in dorsolateral prefrontal cortex

15.

16.

17.

18.

19.

20.

21.

22.

during a working memory task in manic, euthymic and depressed bipolar subjects. Psychiatry Res 2010; 182: 22–29. Delaloye C, Moy G, de Bilbao F et al. Longitudinal analysis of cognitive performances and structural brain changes in late-life bipolar disorder. Int J Geriatr Psychiatry 2011; 26: 1309–1318. Lyoo K, Renshaw PF. Functional magnetic resonance imaging, diffusion tensor imaging, and magnetic resonance spectroscopy in bipolar disorder. In: Yatham LN, Maj M, eds. Bipolar Disorder – Clinical and Neurobiological Foundations. Oxford: Wiley Blackwell, 2010: 231–345. Strakowski SM, Delbello MP, Adler CM. The functional neuroanatomy of bipolar disorder: a review of neuroimaging findings. Mol Psychiatry 2005; 10: 105–116. Yurgelun-Todd DA, Gruber SA, Kanayama G, Killgore WD, Baird AA, Young AD. fMRI during affect discrimination in bipolar affective disorder. Bipolar Disord 2000; 2: 237–248. Adler CM, Holland SK, Schmithorst V, Tuchfarber MJ, Strakowski SM. Changes in neuronal activation in patients with bipolar disorder during performance of a working memory task. Bipolar Disord 2004; 6: 540–549. Drapier D, Surguladze S, Marshall N et al. Genetic liability for bipolar disorder is characterized by excess frontal activation in response to a working memory task. Biol Psychiatry 2008; 64: 513–520. Lagopoulos J, Malhi GS. A functional magnetic resonance imaging study of emotional Stroop in euthymic bipolar disorder. Neuroreport 2007; 18: 1583–1587. Monks PJ, Thompson JM, Bullmore ET et al. A functional MRI study of working memory task in euthymic bipolar disorder: evidence for task-specific dysfunction. Bipolar Disord 2004; 6: 550–564. Oertel-Knochel V, Reinke B, Hornung A et al. Patterns of autobiographical memory in bipolar disorder examined by psychometric and functional neuroimaging methods. J Nerv Ment Dis 2012; 200: 296–304. Deckersbach T, Dougherty DD, Savage C et al. Impaired recruitment of the dorsolateral prefrontal cortex and hippocampus during encoding in bipolar disorder. Biol Psychiatry 2006; 59: 138–146. Brooks JO 3rd, Rosen AC, Hoblyn JC et al. Resting prefrontal hypometabolism and paralimbic hypermetabolism related to verbal recall deficits in euthymic older adults with bipolar disorder. Am J Geriatr Psychiatry 2009; 17: 1022–1029. Arnone D, Cavanagh J, Gerber D et al. Magnetic resonance imaging studies in bipolar disorder and schizophrenia: meta-analysis. Br J Psychiatry 2009; 195: 194–201. Ellison-Wright I, Bullmore E. Anatomy of bipolar disorder and schizophrenia: a meta-analysis. Schizophr Res 2010; 117: 1–12. Nortje G, Stein DJ, Radua J et al. Systematic review and voxel-based meta-analysis of diffusion tensor imaging studies in bipolar disorder. J Affect Disord 2013; 150: 192–200. Versace A, Thompson WK, Zhou D et al. Abnormal left and right amygdala-orbitofrontal cortical functional connectivity to emotional faces: state versus trait vulnerability markers of depression in bipolar disorder. Biol Psychiatry 2010; 67: 422–431. Haller S, Xekardaki A, Delaloye C et al. Combined analysis of grey matter voxel-based morphometry and white matter tract-based spatial statistics in late-life bipolar disorder. J Psychiatry Neurosci 2011; 36: 391–401.

843

Oertel-Kn€ ochel et al. 23. Strakowski SM, Adler CM, Almeida J et al. The functional neuroanatomy of bipolar disorder: a consensus model. Bipolar Disord 2012; 14: 313–325. 24. Anticevic A, Repovs G, Shulman GL et al. When less is more: TPJ and default network deactivation during encoding predicts working memory performance. Neuroimage 2010; 49: 2638–2648. 25. Sestieri C, Corbetta M, Romani GL et al. Episodic memory retrieval, parietal cortex, and the default mode network: functional and topographic analyses. J Neurosci 2011; 31: 4407–4420. 26. APA. Diagnostic and Statistical Manual of Mental Disorders, 4th edn. Washington, DC: American Psychiatric Association, 1994. 27. Wittchen H-U, Wunderlich U, Gruschwitz S et al. Strukturiertes Klinisches Interview f€ ur DSM-IV (SKID). G€ ottingen: Beltz-Test, 1996. 28. Hautzinger M, Keller F, K€ uhner C. Das Beck Depressionsinventar II. Deutsche Bearbeitung und Handbuch zum BDI II. Frankfurt am Main: Harcourt Test Services, 2006. 29. Bech P. Rating scales for affective disorders: their validity and consistency. Acta Psychiatr Scand Suppl 1981; 295: 1–101. 30. Krohne HW, Egloff B, Kohlmann C-W et al. Untersuchungen mit einer deutschen Version der “Positive and Negative Affect Schedule” (PANAS). [Investigations with a German version of the Positive and Negative Affect Schedule (PANAS)]. Diagnostica 1996; 42: 139–156. 31. Franke G. Die Symptom-Checkliste von Derogatis – deutsche Version. G€ ottingen: Beltz Test, 2002. 32. Niemann H, Sturm W, T€ ohne-Otto AIT et al. California Verbal Learning Test (CVLT). Deutsche Adaptation. Frankfurt: Pearson Assessment and Information GmbH, 2008. 33. Lehrl S. Mehrfachwahl-Wortschatz-Intelligenztest M-W-T B. G€ ottingen: Spitta Verlag GmbH, 2005. 34. Reitan RM, Hom J, Wolfson D. Verbal processing by the brain. J Clin Exp Neuropsychol 1988; 10: 400–408. 35. MacMillan N, Creelman C. Detection Theory: A User’s Guide. New Jersey: Lawrence Erlbaum Associations, 2005. 36. Deichmann R, Schwarzbauer C, Turner R. Optimisation of the 3D MDEFT sequence for anatomical brain imaging: technical implications at 1.5 and 3T. Neuroimage 2004; 21: 757–767. 37. Griswold MA, Jakob PM, Heidemann RM et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002; 47: 1202–1210. 38. Talairach J, Tournoux P. Co-planar Stereotexic Atlas of the Human Brain. New York, NY: Thieme Medical, 1988. 39. Boynton GM, Engel SA, Glover GH et al. Linear systems analysis of functional magnetic resonance imaging in human V1. J Neurosci 1996; 16: 4207–4221. 40. Ashburner J, Friston KJ. Voxel-based morphometry – the methods. Neuroimage 2000; 11: 805–821. 41. Maldjian JA, Laurienti PJ, Burdette JB et al. An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Neuroimage 2003; 19: 1233–1239. 42. Smith SM, Jenkinson M, Johansen-Berg H et al. Tractbased spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage 2006; 31: 1487–1505. 43. Chavez S, Storey P, Graham SJ. Robust correction of spike noise: application to diffusion tensor imaging. Magn Reson Med 2009; 62: 510–519. 44. Nichols TE, Holmes AP. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum Brain Mapp 2002; 15: 1–25.

844

45. Hua K, Zhang J, Wakana S et al. Tract probability maps in stereotaxic spaces: analyses of white matter anatomy and tract-specific quantification. Neuroimage 2008; 39: 336–347. 46. Genovese CR, Lazar NA, Nichols T. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage 2002; 15: 870–878. 47. Almeida JR, Mechelli A, Hassel S et al. Abnormally increased effective connectivity between parahippocampal gyrus and ventromedial prefrontal regions during emotion labeling in bipolar disorder. Psychiatry Res 2009; 30: 195–201. 48. Chen CH, Suckling J, Lennox BR, Ooi C, Bullmore ET. A quantitative meta-analysis of fMRI studies in bipolar disorder. Bipolar Disord 2011; 13: 1–15. 49. Houenou J, Frommberger J, Carde S et al. Neuroimagingbased markers of bipolar disorder: evidence from two meta-analyses. J Affect Disord 2011; 132: 344–355. 50. Phillips ML, Ladouceur CD, Drevets WC. A neural model of voluntary and automatic emotion regulation: implications for understanding the pathophysiology and neurodevelopment of bipolar disorder. Mol Psychiatry 2008; 13: 833–857. 51. van Strien NM, Cappaert NL, Witter MP. The anatomy of memory: an interactive overview of the parahippocampal-hippocampal network. Nat Rev Neurosci 2009; 10: 272–282. 52. Schneider MR, DelBello MP, McNamara RK, Strakowski SM, Adler CM. Neuroprogression in bipolar disorder. Bipolar Disord 2012; 14: 356–374. 53. Liu Y, Spulber G, Lehtim€ aki KK et al. Diffusion tensor imaging and tract-based spatial statistics in Alzheimer’s disease and mild cognitive impairment. Neurobiol Aging 2011; 32: 1558–1571. 54. O’Dwyer L, Lamberton F, Bokde AL et al. Multiple indices of diffusion identifies white matter damage in mild cognitive impairment and Alzheimer’s disease. PLoS ONE 2011; 6: e21745. 55. Prvulovic D, Van de Ven V, Sack AT et al. Functional activation imaging in aging and dementia. Psychiatry Res 2005; 140: 97–113. 56. Manoach DS. Prefrontal cortex dysfunction during working memory performance in schizophrenia: reconciling discrepant findings. Schizophr Res 2003; 60: 285–298. 57. Deckersbach T, McMurrich S, Ogutha J et al. Characteristics of non-verbal memory impairment in bipolar disorder: the role of encoding strategies. Psychol Med 2004; 34: 823–832. 58. Behnken A, Schoning S, Gerss J et al. Persistent nonverbal memory impairment in remitted major depression – caused by encoding deficits? J Affect Disord 2010; 122: 144–148. 59. Martinez-Aran A, Torrent C, Sole B et al. Functional remediation for bipolar disorder. Clin Pract Epidemiol Ment Health 2011; 7: 112–116. 60. Weiskopf N. Real-time fMRI and its application to neurofeedback. Neuroimage 2012; 62: 682–692. 61. Keedwell PA, Linden DE. Integrative neuroimaging in mood disorders. Curr Opin Psychiatry 2013; 26: 27–32.

Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. (a) Main effect of encoding across groups. (b) Main effect of retrieval across groups. A = anterior; BA = Brodmann

Episodic memory in bipolar disorders area; LH = left hemisphere; P = posterior; RH = right hemisphere. The activation maps were superimposed on a surfacereconstructed Montreal Neurological Institute (MNI) template. The left side of the figure indicates the right side of the brain (radiological convention). The color codes denote t-scores of all relevant areas. Table S1. Region-of-interest (ROI) group comparisons (t-tests) (thresholded at cluster level a ≤ 0.05%) between controls (n = 20) and BP patients (n = 21) during non-verbal episodic memory encoding and retrieval, using a bilateral hippocampus mask provided by BrainVoyager QX (48,631 lL; centers of mass: CON = healthy controls; LH: x = 20; y = 30; z = 4;

RH = right hemisphere: x = 22; y = 29, z = 2). BA = Brodmann area; MNI = Montreal Neurological Institute (MNI) coordinates; PAT = bipolar patients; ns = not significant; TAL = Talairach coordinates. **p < 0.01; *p < 0.05. Table S2. Correlation analysis between areas which showed significant group differences during memory-related functional magnetic resonance imaging and verbal episodic memory performance across groups. CON = healthy controls; CVLT YN = California Verbal Learning Test, Yes/No retrieval; NV = accuracy in the non-verbal episodic memory task; PAT = bipolar patients.

845

Copyright of Bipolar Disorders is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

Episodic memory impairments in bipolar disorder are associated with functional and structural brain changes.

We combined multimodal functional magnetic resonance imaging (fMRI) and structural magnetic resonance imaging to probe abnormalities in brain circuits...
939KB Sizes 0 Downloads 7 Views