2 Methods

2.1 Participants

The group of 68 participants consisted of 23 healthy controls with no history of psychiatric disease (HC), 20 patients with a current diagnosis of MDD (MDD), and 25 patients with a current comorbid diagnosis of MDD and at least one anxiety disorder (10 patients with agoraphobia, 6 with social phobia, 13 with panic disorder and 7 with generalized anxiety disorder; 8 patients exhibited more than one AD at the time of testing) (MDD+AD). Patients were recruited from the Clinic of Psychiatry and Psychotherapy at the University Hospital Magdeburg and affiliated mental health services in the area. Healthy controls were recruited from the local community and matched for age and sex. Table 1 shows participant characteristics split by participant group. The mean age of the participants was 35.25 (12.4) years. 52.9 % of the participants were male and the mean years of education was 15.5 (3.4). Of the patient groups, 73.4 % were currently receiving psychopharmacotherapy and the mean duration of illness was 57.14 months.
Exclusion criteria were: (1) age under 18 or over 65, (2) a history of other comorbid psychiatric disorders, substance abuse or dependence, (3) other severe medical illness, head injury or neurological disorders, (4) MRI contraindications. Demographic variables and inclusion/exclusion criteria were assessed using a standardized questionnaire and a structured interview conducted by trained medical staff. Diagnoses were confirmed according to DSM-IV (American Psychiatric Association, 2013) criteria using the Structured Clinical Interview for DSM-IV (SCID I and II) (First et al., 1996). In addition, the Beck Depression Inventory II (BDI-II) (Beck et al., 1996) and the Beck Anxiety Inventory (BAI) (Beck et al., 1988) were filled out by each participant as a measure of symptom severity. Neurological disorders were examined via a neuropsychological test battery.
All participants received detailed oral and written information about the study and gave written informed consent prior to participation. Ethical approval for the study was granted by the Ethics Committee of the Otto-von-Guericke University. The study was conducted in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki).

2.2 MRI data acquisition

Magnetic resonance images were obtained using a 3 Tesla Siemens MRI scanner (MAGNETOM Prisma Syngo MR D13D; Siemens, Erlangen, Germany) with a 64-channel phased array head coil. All subjects received earplugs to protect them from the noise of the head coil. Structural images were recorded using a magnetization-prepared rapid gradient echo (MPRAGE) sequence. High-resolution T1-weighted three-dimensional anatomical scans of 192 sagittal slices with a voxel size of 1.0 x 1.0 x 1.0 mm3 of the whole brain were obtained (repetition time (TR): 2500 ms, echo time (TE): 2.82 ms, inversion time (TI): 1100 ms, field of view (FOV) from foot to head (FH): 256 mm, anterior to posterior (AP): 256 mm and right to left (RL): 192 mm, flip angle of 7°, matrix size = 256 x 256). The scan time required for structural acquisition was 9 min 20 s.
T2-weighted functional images were collected in single runs using echo-planar imaging (TE = 30 ms; TR = 2000 ms; field of view = 216 mm; flip angle = 90°, voxel size 3 × 3 × 3 mm, EPI-Factor 72) using blood oxygenation level-dependent contrast. Covering the whole brain, 36 contiguous 3-mm-thick slices were acquired sequentially in ascending order parallel to the anterior–posterior-commissure. Within a total scan time of 10:06 min, 306 volumes were acquired continuously. Participants were instructed to stay awake with their eyes closed during the resting-state scans.

2.3 MRI data pre-processing

Functional data were preprocessed using Data Processing Assistant for Resting-State fMRI (DPARSF 4.3) (Chao-Gan & Yu-Feng, 2010). Imaging preprocessing included: interleaved slice time correction, realignment to temporal mean image, motion regression, normalization to Montreal Neurological Institute (MNI) reference space combined with reslicing to an isotropic resolution of 3-mm, temporal band-pass filtering (0.01 < f < 0.10 Hz) and smoothing using a full-width-at-half-maximum Gaussian-kernel of 4 mm. Spurious variance was removed by nuisance regression including covariates for head motion using the Friston 24-parameter model as well as averaged signals from global brain signal, cerebrospinal fluid and white matter signal.
To compute RSFC within and between the three networks, we proceeded as follows. First, for each network, we utilized established a priori defined, literature-based seed regions (Table 2) belonging to the DMN, VAN and ECN to compute seed-to-whole-brain connectivity maps by calculating the Pearson correlation coefficient of the average time series of each seed region to the rest of the brain. Then, we converted the correlation coefficient maps using Fisher’s z-transformation to improve the Gaussianity of their distribution.

2.4 Statistical MRI data analysis

Following preprocessing of the data, using Statistical Parametric Mapping (SPM 12) (Friston, 2007) functional connectivity maps were compared using analyses of variance, where age and sex were included as covariates. For each network-pair, we then computed group-wise comparisons between all the three groups using post-hoc t-tests for independent samples. All clusters were formed using a cluster-forming-threshold of p < .001 and controlled for Family-Wise-Error (pFWE-corr < .05). Additonally, all clusters were subsequently corrected for multiple comparisons using Bonferroni-correction and considered significant atpcorr < .05. The atlas of Yeo et al. (2011) was used to assign the resulting clusters to one of the three networks. For all 5 seeds, the within-network (seed to other structures within the network the seed belonged to) and the between-network connectivity (seed to structures belonging to the other networks included in this study) was examined. For post-hoc analysis using SPSS 27.0 (IBM Corp., 2022), we extracted the beta values of these clusters indicating the degree of connectivity from our first level models using the REX toolbox (Gabrieli Lab). Automated anatomical labelling atlas (AAL 3) (Rolls et al., 2020) was used to identify and label clusters with aberrant RSFC. Connectivities were visualized using Brain Net Viewer (2013).
SPSS (IBM Corp., 2022) software was used for comparison of group characteristics as well as correlation and regression analysis. All results were considered to be significant atpcorr < .05 after correcting for multiple comparisons using Bonferroni-correction. Characteristics of the participant groups were summarized using mean and standard deviation for continuous variables and percentage for categorial variables. Variables were tested for normal distribution using Shapiro-Wilk-Test. Age and clinical scores were compared between all groups using ANOVA. Between patient groups duration of illness and clinical scores were compared using Mann-Whitney-U -test. Differences in the distribution of sex and medication were assessed using χ2 -tests.
To further investigate the nature of the relationship of RSFC-values and symptom severity, we performed regression analyses using linear models, for all clusters found to be significant within the patient group (MDD and MDD+AD). Mean beta values were used as dependent variables and severity of symptoms as measured by BDI-II and BAI-scores as independent variables. Additionally, since an effect of pharmacotherapy on RFSC has commonly been found (Rolls et al., 2019), we performed an independent-sample-t-test within the patient group, testing for a difference in RSFC-values between medication groups. Bonferroni-correction was used to correct for multiple comparisons.