A total of 42 right-handed, currently depressed individuals diagnosed with BD II were recruited from the psychiatry department, First Affiliated Hospital of Jinan University, Guangzhou, China. The patients were aged from 18 to 55 years. All patients met Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (known as DSM-V) criteria for BD II according to the diagnostic assessment by the Structured Clinical Interview for DSM-V Patient Edition (SCID-P) by two experienced psychiatrists (Y.J. and S.Z., with 21 and 6 years of experienced clinical psychiatry, respectively). The clinical state was assessed by using the 24-item Hamilton Depression Rating Scale (HDRS) and the Young Mania Rating Scale (YMRS) during the 3-day period prior to the imaging session. The inclusion criterion for the depressed patients with BD II was a total HDRS-24 score > 21 and YMRS score < 7. The exclusion criteria were patients with other Axis-I psychiatric disorders, a history of electroconvulsive therapy, neurological disorders, any history of organic brain disorder, mental retardation, pregnancy, alcohol/substance abuse, cardiovascular diseases or any presence of a concurrent and major physical illness. At the time of testing, all patients were either medication-naïve, or were not medicated for at least six months. In addition, 69 right-handed HC were recruited via local advertisements. They were carefully screened through a diagnostic interview, the Structured Clinical Interview for DSM-V Nonpatient Edition (SCID-NP), to rule out the presence of current or past history of any psychiatric illness. Further exclusion criteria for HCs were any history of psychiatric illness in first-degree relatives, current or past significant medical or neurological illness.
2.2 MR Imaging Data Acquisition and Preprocessing
All MRI data were obtained on a GE Discovery MR750 3.0T System with an eight-channel phased-array head coil. The participants were scanned in a supine, head-first position with symmetrically placed cushions on both sides of the head to decrease motion. During the scanning, the participants were instructed to relax with their eyes closed without falling asleep. After the experiment, each participant confirmed not having fallen asleep.
The rs-fMRI data were acquired using a gradient-echo echo-planar imaging sequence with the following parameters: time repetition (TR)/time echo (TE) = 2000/25 ms; flip angle = 90°; voxel size = 3.75 × 3.75 × 3 mm³; field of view (FOV) = 240 × 240 mm2; matrix = 64 × 64; slice thickness/gap = 3.0/1.0 mm; 35 axial slices covering the whole brain; and 210 volumes acquired in 7 min. In addition, a three-dimensional brain volume imaging (3D-BRAVO) sequence covering the whole brain was used for structural data acquisition with the following parameters: TR/TE = 8.2/3.2 ms; flip angle = 12°; bandwidth = 31.25 Hz; slice thickness/gap = 1.0/0 mm; matrix = 256 × 256; FOV = 240 × 240 mm; NEX = 1; and acquisition time = 3 min 45 s. Routine MRI examination images were also collected for excluding any anatomic abnormality. All participants were found by two experienced neuroradiologists (YM and YS, with 8 and 3 years of experience in neuroimaging, respectively) to confirm the absence of any brain structural abnormalities.
2.3 Functional Image Preprocessing
The preprocessing was carried out using Data Processing Assistant for Resting-State fMRI (DPABI_V3.0, http://restfmri.net/forum/DPABI)  which is based on Statistical Parametric Mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm/). For each subject, the first 10 images of the rs-fMRI dataset were discarded to ensure steady-state longitudinal magnetization. The remaining 200 images were first slice-time corrected and then were realigned to the first image for correcting for inter-TR head motion. This realignment correction provided a record of the head motion within the rs-fMRI scan. All subjects should have no more than 2 mm maximum displacement in any plane, 2° of angular motion as well as 0.2 mm in mean frame-wise displacement (FD) . The individual T1 structural images were segmented (white matter, gray matter, and cerebrospinal fluid) using a segmentation toolbox. Then, the DARTEL toolbox was used to create a study specific template for the accurate normalization. Then, resting-state functional images were coregistered to the structural images and transformed into standard Montreal Neurological Institute (MNI) space, resliced to a voxel size of 3×3×3 mm³ resolution and smoothed using a 6 mm full width at half maximum (FWHM) Gaussian kernel. The data were removed linear trend and passed through band-pass filtered of 0.01-0.1 Hz.
2.4 Component selection
Following preprocessing, images were processed in the Group ICA FMRI Toolbox (GIFT) (http://icab.sourceforge.net). Data were first prewhitened and dimensions reduced via a 2-step principal component analysis . Estimation showed an estimate of 82 components. Then, images of all subjects were decomposed into a set of 82 spatially independent components by the Infomax algorithm. Each independent component (IC) depicted a distinct network of brain regions that have the same pattern of homodynamic change over time, and robustness of the component was achieved by running ICASSO  (http://www.cis.hut.fi/projects/ica/icasso) for 100 iterations. We then chose the components using component labeler in GIFT. The component labeler option is provided to label components given the templates of interest. Each component is correlated with the given templates and best template is selected based on the maximum correlation value. The 7-network templates used for sorting were created by yeo (https://surfer.nmr.mgh.harvard.edu/fswiki/CorticalParcellation_Yeo2011), and the networks including Visual Network, Somatomotor Network, Dorsal Attention Network, Ventral Attention Network, Limbic Network, Frontoparietal Network, and Default Network. After this procedure, nine and eight of components were respectively identified as components of the Somatomotor Network and Limbic Network.
Individual maps of each group were subjected to a random effect one sample t-test in SPM12. Deemed significant at uncorrected p<0.01 with a minimum extent threshold of 10 contiguous voxels and created a group-specific component map. These two maps of BD patients and HCs were combined as a mask for group analyses.
2.5 Pro-inflammatory Cytokines Measures
Blood samples from BD patients and HCs were obtained in the morning under fasting condition, abstained from alcoholic beverages for at least one day prior to testing and processed (then frozen) by technicians. Fasting serum samples were collected in serum tubes, clotted for 30 min, and stored at -80℃ until use. Levels of pro-inflammatory cytokines, including IL-6 and IL-8 levels were determined from serum by the Bio-Plex Pro Human Cytokine Assay kit (Bio-Rad). According to manufacturer’s directions using a Bio-Plex 200 array reader (Bio-Rad). Bio-Plex Manager Software, version 6.1, was used for data acquisition (Bio-Rad).
2.6 Group comparison and Correlation analyses
Independent-sample t-test (normal variable) and Mann-Whitney U test (skewed variables) were used to compare demographic data (except gender) and levels of serum pro-inflammatory cytokines between the two groups with SPSS 19.0 software (SPSS, Chicago, IL, USA). A chi-squared test was performed to compare gender distribution. All tests were two-tailed, and P < 0.05 was considered statistically significant. The two-sample t-test was performed to assess the significant differences of the interest component between BD II patients and HCs within the union mask of one-sample t-test results of both groups. Age, gender, years of education and the mean FD were included as nuisance covariates in the comparisons. Statistical maps were thresholded using permutation tests (PTs) as implemented in PALM and integrated into DPABI. The threshold-free cluster enhancement (TFCE) and voxel wise correction (VOX) with PT were tested at two-tailed P < 0.05 for multiple comparisons. The number of permutations was set at 1000. Previous study observed that permutation test with TFCE, a strict multiple comparison adjustment strategy, reached the best balance between family-wise error rate (under 5%) and test-retest reliability and replicability .
When statistically significant group differences were observed in brain regions and pro-inflammatory cytokines levels, partial correlation analysis was used to compute the correlation between FC values and inflammatory cytokines levels in BD II depression whilst controlling for the effect of gender, age and years of education. Also, the partial correlation coefficients were calculated between the clinical variables and abnormal FC values, abnormal inflammatory cytokines levels in BD II group. These clinical variables included onset age of illness, number of episodes, duration of illness, 24-item HDRS scores and YMRS scores.