Association between resting-state functional connectivity of amygdala subregions and peripheral pro-inflammation cytokines levels in bipolar disorder

The pathophysiological mechanisms of bipolar disorder (BD) are not completely known, and systemic inflammation and immune dysregulation are considered as risk factors. Previous neuroimaging studies have proved metabolic, structural and functional abnormalities of the amygdala in BD, suggesting the vital role of amygdala in BD patients. This study aimed to test the underlying neural mechanism of inflammation-induced functional connectivity (FC) in the amygdala subregions of BD patients. Resting-state functional MRI (rs-fMRI) was used to delineate the amygdala FC from two pairs of amygdala seed regions (the bilateral lateral and medial amygdala) in 51 unmedicated BD patients and 69 healthy controls (HCs). The levels of pro-inflammatory cytokines including interleukin (IL)-1β, IL-6 and tumor necrosis factor (TNF)-α were measured in the serum. The correlation between abnormal levels of pro-inflammatory cytokines and FC values were calculated in BD patients. The BD group exhibited decreased FC between the right medial amygdala and bilateral medial frontal cortex (MFC), and decreased FC between the left medial amygdala and the left temporal pole (TP), right orbital inferior frontal gyrus compared with HCs. The BD patients had higher levels of TNF-α than HCs. Correlation analysis showed negative correlation between the TNF-α level and abnormal FC of the right medial amygdala-bilateral MFC; and negative correlation between TNF-α levels and abnormal FC of the left medial amygdala-left TP in BD group. These findings suggest that dysfunctional and immune dysregulation between the amygdala and the frontotemporal circuitry might play a critical role in the pathogenesis of BD.


Introduction
Bipolar disorder (BD) is a chronic, recurrent and disabling psychiatric illness. It is a leading cause of disability in young people as they can lead to functional and cognitive impairment and reduce quality of life and increase mortality (Kupfer, 2005;Vieta et al., 2018). Early and accurate diagnosis of BD is difficult in clinical practice, and it is often misdiagnosed as unipolar depression as the onset of BD is commonly characterized by nonspecific symptoms, mood lability or a depressive episode. Although it has been studied for over decades, the pathophysiology of BD remained unclear. Dysfunction of innate immune system that leads to neuroinflammation has been considered as risk factor in the pathophysiology of numerous psychiatric illnesses including BD (Brunoni et al., 2020;Dickerson et al., 2017;Goldsmith et al., 2016;Lin et al., 2020).
Jiaying Gong and Guanmao Chen contributed equally to this work.
Resting-state functional magnetic resonance imaging (rs-fMRI) is a primary noninvasive neuroimaging technique that can reflect the changes of brain function status and resting state functional connectivity (rs-FC), and is a fundamental tool in reflecting the correlations of activities between different brain regions, helping to understand brain regions with altered connectivity and function in mental disorders (Deco & Kringelbach, 2014;Mulders et al., 2015). Previous neuroimaging studies have reported widespread structural and functional abnormalities in brain region of the amygdala, including reduced gray matter volume (Yu et al., 2019), broken white matter integrity, decreased FC (Johnston et al., 2017) and lower activation (Jesudas et al., 2019), ais associated with the pathophysiology of BD. The amygdala is a complex consisting of structural and functional heterogeneous nuclei, tucked in the medial temporal lobe and comparatively small in size. It plays a central role in regulating emotions and mediating fear responses (Roy et al., 2009;Wu et al., 2016). The amygdala complex is grouped into three distinct subregions cytoarchitectonically: centromedial, basolateral, and superficial (Jalbrzikowski et al., 2017;Wu et al., 2016). As there several studies that discussed different functions of centromedial (including the central and medial nuclei) and basolateral (including the lateral, basolateral, basomedial, and basoventral nuclei) amygdala, but not the superficial region (Jalbrzikowski et al., 2017;Mosher et al., 2010;Pape & Pare, 2010), this study focused on the centromedial and basolateral subregions that are similar to the medial and lateral masks according to the Brainnetome atlas (Jalbrzikowski et al., 2017). The lateral amygdala receives sensory information and involves assessment of emotional content of sensory information, while the medial amygdala is regarded as the major output station of the entire amygdala and involves enhancement of attention allocation and determine the salience of input (Bzdok et al., 2013;Janak & Tye, 2015;Roy et al., 2009). Existing studies have put forwarded the changes in amygdala FC that are related to BD accompanied by the frontal lobe (including the dorsolateral prefrontal cortex (Zhang et al., 2018a), prefrontal cortex , and inferior frontal gyrus [IFG] (Ho et al., 2019)), supplementary motor area  and anterior hippocampus (Ambrosi et al., 2017). Among these studies, only one study used the amygdala subregions FC, but through region of interest (ROI)-to-ROI analyses instead of whole-brain approach based on subregions seed masks of each subregion (Ambrosi et al., 2017). However, the functional connectivity between the amygdala subregions and the brain areas has not been examined in detailed manner in BD.
Cytokines act as regulators of host responses to infection, inflammation, trauma, and immune responses (Dinarello, 2000). During the process of inflammation, proinflammatory cytokines access the central nervous system and interact with cytokine network in the brain to virtually influence each and every aspect of brain function that is relevant to the behavior including neurotransmitter metabolism, neuroendocrine function, synaptic plasticity, mood regulation, motor activity, motivation, anxiety and panic (Capuron & Miller, 2011). Immune system dysfunction leads to the occurrence of neuroinflammation in BD and increasingly implicated in the pathophysiology of this mental disorder (Najjar et al., 2013;Rosenblat & McIntyre, 2016). Several studies have shown elevated proinflammatory cytokines including but not limited to IL-6, IL-1β, and TNF-α in patients with BD (Brunoni et al., 2020;Pan et al., 2020), particular in acute mania (Brietzke et al., 2009;Fiedorowicz et al., 2015;Isgren et al., 2017;Modabbernia et al., 2013) and depressive episodes (Brietzke et al., 2009;Fiedorowicz et al., 2015;Isgren et al., 2017;Modabbernia et al., 2013). Previous metaanalyses have reported that IL-1β, IL-6, and TNF-α are the most common abnormal cytokines in BD patients (Goldsmith et al., 2016;Modabbernia et al., 2013;Rowland et al., 2018). Some MRI studies have revealed abnormal structure and dysfunction in specific brain regions (such as amygdala, IFG, medial frontal cortex [MFC], and temporal lobe [TP]) and its correlation with IL-1β, IL-6, and TNF-α in the peripheral area, and might contribute to the symptoms of fear, anxiety and emotional disturbances. This might help us to better understand the role of IL-1β, IL-6, and TNF-α, and its relation to the pathophysiology of mood disorder (Chen et al., 2019b;Felger, 2018;Fiedorowicz et al., 2015;Wilczynska et al., 2018). A rs-fMRI study revealed that higher IL-6 receptor levels are associated with connectivity abnormalities between the medial prefrontal cortex and amygdala involved in BD (Tu et al., 2017).
In our study, the whole-brain FC of each amygdala subdivision in BD patients and HCs was calculated, and the levels of IL-1β, IL-6, and TNF-α were obtained in the peripheral blood of these participants. This study aimed to explore the interrelations between pro-inflammatory cytokines and amygdala subdivisions FC in patients with BD. We hypothesized that both the amygdala subdivisions FC and pro-inflammatory cytokines levels are abnormal, and the amygdala dysfunction is linked to the changes in cytokines.

Participants
A total of 55 right-handed individuals diagnosed with BD (15 BD I and 40 BD II) were recruited from the in-and outpatients of psychiatry department, First Affiliated Hospital of Jinan University, Guangzhou, China. The patients' age ranged from 18 to 55 years. All patients met the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (known as DSM-V) criteria for BD according to the diagnostic assessment by the Structured Clinical Interview for DSM-V Patient Edition (SCID-P) put forwarded by two experienced psychiatrists (Y.J. and S.Z., with 21 and 6 years of experienced clinical psychiatry, respectively). The clinical state of BD patients 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 exclusion criteria included patients with other Axis-I psychiatric disorders (except for BD II and anxiety symptoms) as assessed by two experienced psychiatrists (Y.J. and S.Z., with 21 and 6 years of experienced clinical psychiatry, respectively) through SCID-P, a history of electroconvulsive therapy, other neurological disorders, any history of organic brain disorder, autoimmune/immune diseases, mental retardation, pregnancy, alcohol/substance abuse, cardiovascular diseases or presence of any concurrent and major physical illnesses as reported by laboratory examination or self-reported by patients in the experienced psychiatrists' interview and when conversing with patients. Those patients who currently took anti-inflammatory medications or immunomodulators (i.e., statins or metformin) were excluded from the study. At the time of testing, all patients were either medication-naive, or were not medicated for at least six months. The remaining patients were receiving mood stabilizers (lithium, sodium valproate) and/ or antidepressants (duloxetine or paroxetine). In addition, 70 right-handed healthy controls (HCs) 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 psychiatric illnesses. Further exclusion criteria for HCs include history of psychiatric illnesses in first-degree relatives, current or past significant medical or neurological illnesses.
This study was approved by the Ethics Committee of First Affiliated Hospital of Jinan University, Guangzhou, China. All subjects were right-handed and signed a written informed consent form after full writing and verbal explanation of the study. Two senior clinical psychiatrists have confirmed that all subjects had the ability to participate in the examination.

MR imaging data acquisition and preprocessing
All MRI data were obtained by using a GE Discovery MR 750 3.0T System with an 8-channel phased array head coil. Subjects were scanned in a supine, head-first position with symmetrically placed cushions on both sides of the head to decrease motion. During the scanning, participants were instructed to relax with their eyes closed without falling asleep; and after the experiment, each participant confirmed not having fallen asleep.

Functional image preprocessing
The preprocessing was carried out using the Data Processing Assistant for Resting-State fMRI (DPABI_V3.0, http:// restf mri. net/ forum/ DPABI) (Yan et al., 2016), and this 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 slice-time corrected initially and then were realigned to the first image for correcting the inter-TR head motion. This realignment correction provides a record of the head motion within the rs-fMRI scan. All subjects should not have more than 2 mm maximum displacement in any plane, 2° of angular motion as well as 0.2 mm in the mean frame-wise displacement (FD) (Jenkinson et al., 2002). The individual T1 structural images were segmented (white matter, gray matter, and cerebrospinal fluid) using a segmentation toolbox. The DARTEL toolbox was then used to create a study specific template for accurate normalization. The restingstate functional images were co-registered with the structural images and transformed into the standard Montreal Neurological Institute (MNI) space, resliced to a voxel size of 3 × 3 × 3 mm³ resolution and smoothened using a 6 mm full width at half maximum (FWHM) Gaussian kernel. The data were removed the linear trend and passed through the bandpass filter of 0.01-0.1 Hz. The global mean signal regression is commonly used in the field of rs-fMRI, and it could afford decreased dependencies on head motion (Yan et al., 2013a) and increased tissue sensitivity (Fox et al., 2009). The signals from white matter and cerebrospinal fluid were regressed out to reduce both respiratory and cardiac effects (Yan et al., 2013b). The whole brain functional connectivity estimates were influenced by head motion and the head motion could be misinterpreted as neuronal effects easily (Van Dijk et al., 2012). Several spurious covariates and their temporal derivatives were then regressed out from the time course of each voxel, including the signals of the brain global mean, white matter, and cerebrospinal fluid as well as the Friston-24 parameters of head motion (including the 6 head motion parameters, 6 head motion parameters at one time point before, and the 12 corresponding squared items) (Friston et al., 1996).

Resting-state functional connectivity
Consistent with the previous studies of amygdala FC (Jalbrzikowski et al., 2017), four seed masks (the bilateral lateral amygdala and medial amygdala) were selected using Brainnetome atlas (http:// atlas. brain netome. org/ bnatl as. php) (Fan et al., 2016). The time course of the masks was correlated against all other voxels within the whole brain. Individual rs-FC maps of the lateral amygdala, and medial amygdala were generated by calculating the Pearson's correlation coefficients between the mean time series of the masks and the time series of each voxel in the whole brain. The subject-level correlation maps were then converted to the z-value maps using Fisher's transformation to improve the normality. For all the subjects, four z-score maps that represent the intrinsic FC of the four amygdala masks were finally obtained.

Pro-inflammatory cytokines measures
Blood samples from BD patients and HCs were obtained in the morning under fasting condition, followed by abstaining by alcoholic beverages for at least one day prior to testing and then were processed (then frozen) by the technicians. Through venipuncture, four milliliters of fasting blood were drawn from each subject into a sterile vacuum tube (Becton & Dickinson 367,812; Becton Dickinson, Franklin Lakes, NJ, USA), clotted for 30 min, and then centrifuged at 1,000 g for 15 min at 4 ℃. The serum was centrifuged again (3,000 rpm for 10 min at 4 ℃), and stored at -80 ℃ until use. Before testing, the serum should be melted on ice at low temperature, and the testing process was carried out at room temperature in accordance with the requirements of the kit to ensure effective combining of antigen and antibody. The levels of pro-inflammatory cytokines, including IL-6, IL-1β, and TNF-α were determined from the serum by the Bio-Plex Pro Human Cytokine Assay kit (Bio-Rad) according to the 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). The intra-assay coefficients of variation (CVs) were 3.58% for IL-1β, 2.97% for IL-6, and 3.66% for TNF-α. A reliable detection range was provided with the lower limit of quantitation (LLOQ) and the upper limit of quantitation (ULOQ). The detection ranges (LLOQ-ULOQ) were 0.057-1027.514 pg/ml for IL-1β, 0.11-1997.01 pg/ml for IL-6, and 0.69-13336.72 pg/ml for TNF-α. No immune variables were below the limits of assay detection.

Statistical analysis
Independent-sample t-test (normal variable) and Mann-Whitney U test (non-normal variable) were used to compare the demographic data (except gender) and levels of serum proinflammatory 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 to be statistically significant. One-sample t-test was performed on z-score maps for each mask to demonstrate within-group FC spatial distribution of each seed for BD patients and HCs within a brain mask (Fig. S1), and the significant level was set at p < 0.05 (uncorrected). Then the two-sample t-test was performed to assess significant differences of the whole brain FC in each region between BD 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 whole brain FC and cytokine group comparisons. Gaussian random field (GRF) theory was used for cluster-level multiple comparison correction (voxel p value < 0.001; cluster p value < 0.05, GRF corrected). When statistically significant group differences were observed in the brain regions and pro-inflammatory cytokines levels, Spearman correlation analysis was performed to compute the correlation between FC values and pro-inflammatory cytokine levels both in BD and HC groups. Also, Spearman correlation coefficients were calculated between the clinical variables and abnormal FC values, and abnormal pro-inflammatory cytokines levels in the BD group. These clinical variables included onset age of illness, number of episodes, duration of illness, 24-item HDRS scores and YMRS scores. Further, multiple linear regression was adopted to model the relationship between FC values and pro-inflammatory cytokines levels after adjusting for potential confounders including demographic variables, number of cigarettes, body mass index (BMI), and BD subgroup variable ("1" for BD I and "2" for BD II). The significant level was set as p < 0.05. Furthermore, a moderation analysis was performed to clarify the independent and interactive effects of diagnosis and cytokines. Exploring the different correlation between BD and HC would be further performed when there were significant interactions (Supplementary materials).

Demographic and clinical characteristics
The demographic and clinical data of all study participants were presented in Table 1. Four patients in BD (three BD I and one BD II) group and one HC were excluded from further analyses because of excessive head motion. Finally, 51 patients with BD (12 BD I and 39 BD II) and 69 HCs were included in the analysis. For the patients, 39 were in depressed mood state and 12 were in euthymic mood state. The 24-item HDRS score was 23.14 (9.94), and the YMRS score was 8.57 (9.13) for the BD group. There were no significant differences between the BD group and the HCs group in gender, age, BMI, or FD parameters (p > 0.05).

FC analysis within-group
One-sample t-test displayed the within-group FC patterns in the BD and HCs groups. The one-sample t-test revealed that the bilateral lateral amygdala subregions with high FC values were mainly connected to the cerebellum, parahippocampal gyrus, insula, putamen, superior/middle/inferior temporal gyrus, TP, superior/middle/inferior frontal gyrus, cingulate gyrus, pre/postcentral gyrus, supplementary motor area and parietal lobe in the HCs group. The bilateral medial amygdala subregions with high FC values were mainly connected to the parahippocampal gyrus, inferior/middle/superior temporal gyrus, insula, putamen, parietal lobe, cingulate gyrus, MFC, IFG, TP, pre/postcentral gyrus, supplementary motor area, precuneus, cingulate gyrus and cerebellum in the HCs group. The FC spatial distribution in the BD group was similar to that of the HCs group with visual inspection (Fig. 1).

FC changes between-group
Decreased FC between the right medial amygdala and the bilateral MFC and decreased FC between the left medial amygdala and the left TP, the right orbital IFG were observed in the BD group when compared to the HC group. There was no aberrant FC between the right/left lateral amygdala and the cerebral regions in the BD group was observed when compared to the HC group (Table 2; Fig. 2).

Group differences of pro-inflammatory cytokines levels
The levels of IL-6, IL-1β, and TNF-α of the two groups were shown in Table 1. The measurements of serum pro-inflammatory cytokines levels were completed in 51 BD patients and 69 HCs. No significant differences were observed in sex, age, and FD parameter between the two groups. Compared with HCs, BD patients showed increased TNF-α levels (p < 0.05). No statistical differences of IL-6 and IL-1β were observed between the two groups (p > 0.05) ( Table 1).

Correlation analyses
For BD patients, the FC values in the right medial amygdalabilateral MFC showed a negative correlation with TNF-α (r = -0.314, p = 0.025) level; and the FC values in the left medial amygdala-left TP showed a negative correlation with TNF-α (r = -0.330, p = 0.018) level (Fig. 3). For HC group, For the correlations between FC values and clinical variables, including onset age of illness, number of episodes, duration of illness, 24-item HDRS score, and YMRS score, and between levels of TNF-α and clinical variables in BD patients, no significant correlation of survival was observed when Spearman correlation analysis was performed (p > 0.05).
After adjusting for demographic variables, number of cigarettes, BMI, and BD subgroup variable in BD group, the TNF-α showed a negative correlation with the FC values in the right medial amygdala-bilateral MFC (beta = -0.302, t = -2.184, p = 0.032), and a negative association with the FC values in the left medial amygdala-left TP (beta = -0.395, t = -2.694, p = 0.011). Age, number of cigarettes, BMI, and BD subgroup variable showed no association with TNF-α, the FC values in the right medial amygdala-bilateral MFC, and the FC values in the left medial amygdala-left TP (all p > 0.05). In addition, there were no significant relationships between FC values and pro-inflammatory cytokines levels in the HC group. A moderation analysis between inflammatory marker and significant decreased FC in BD and HCs was showed in Table S1.
In addition, the SNR and spatial coverage of the seed amygdala, bilateral MFG, and left TP of each individual's echo planar image were checked in the Table S2 and Fig. S2.   Fig. 1 The FC patterns of the lateral amygdala, medial amygdala in the BD patients and HCs (p < 0.05, uncorrected). The color bar represents functional connection. The T value indicated the intensity of the significant difference which was obtained by one-sample t-test. FC, functional connectivity; BD, bipolar disorder; HCs, healthy controls

Discussion
We performed whole brain amygdala subregions rs-FC and measured the levels of pro-inflammatory cytokines including IL-6, IL-1β, and TNF-α in the serum of BD patients and HCs. The primary findings of this study were as follows: (i). decreased FC was observed between the right medial amygdala and the bilateral MFC, and decreased FC between the left medial amygdala and the left TP, the right orbital IFG in BD group; (ii). increased TNF-α levels were observed in BD patients than that in HCs; (iii). negative correlation was observed between the TNF-α levels and abnormal FC in the right medial amygdala-bilateral MFC; and negative correlation between the TNF-α levels and abnormal Fig. 2 The significant FC differences between the two groups for amygdala seed, respectively (voxel p < 0.001, cluster p < 0.05, GRF corrected). FC, functional connectivity; GRF, Gaussian random field; L (R), left (right) hemisphere FC in the left medial amygdala-left TP was observed in BD patients. To the best of our knowledge, this is the first study to assess the association between the amygdala subregion rs-FC and inflammatory cytokines levels in BD patients, and this assists us to raise awareness with regard to BD from the perspective of intrinsic brain activity and immune system. In the present study, patients with BD showed decreased FC between the right medial amygdala and the bilateral MFC, suggesting a disturbed connectivity between the amygdala and MFC in BD patients. The MFC has been assumed to play a role in emotional regulation by generating preventive responses preceding the emotional events (Ochsner et al., 2009). Previous rs-fMRI studies have shown decreased FC of the MFC and IFG in patients with BD, and this has shed light on the underlying pathophysiology and presentation of impaired emotion regulation and cognitive control (Chen et al., 2019a;Mukherjee et al., 2016;Wang et al., 2016;Zhong et al., 2019). Furthermore, another rs-fMRI study using voxel-mirrored homotopic connectivity (VMHC) has investigated that lower VMHC in the MFC in BD patients. VMHC, is a common rs-fMRI method, that measures the functional connection between each voxel in one hemisphere and the mirror voxel in the other hemisphere, suggesting substantial impairment of interhemispheric coordination in BD patients (Wang et al., 2015). Study of regional function in BD patients showed decreased activity in the MFC (Yu et al., 2017) and IFG regions (Tsujii et al., 2018). Recent task-based fMRI investigation found that patients with BD had neural dysfunction in the MFC, implicating appraisal and expression of emotions (Tseng et al., 2016). Vai et al. have reported that successful antidepressant chronotherapeutics can enhance fronto-limbic connectivity (MFC-amygdala) in emotional processing neural responses in BD patients (Vai et al., 2015). Postmortem studies have revealed dendritic atrophy, oligodendrocyte loss, and gray matter loss in the MFC in BD patients, raising the possibility of abnormal synaptic interactions between the mPFC, amygdala, and OFC among other regions. This may contribute to the disturbances in emotional regulation characteristics of BD (Rajkowska et al., 2001;Savitz et al., 2014). Taken together, our findings of decreased FC between the right medial amygdala and the bilateral MFC might show an abnormal brain integration in default processing of emotional information, displaying the clinical sign of emotional dysregulation in patients with BD.
The TP covering the anterior-most part of the temporal lobe, like a cap lies between the orbital frontal cortex and the amygdala, receives and sends the connections to both regions, playing an important role in socioemotional processing (Olson et al., 2007). In our study, decreased FC was found between the left medial amygdala and the left TP, suggesting disrupted connectivity between the amygdala and the TP. Another fMRI study of BD also exhibited decreased FC from the amygdala to the TP, and this was consistent with our results (Vai et al., 2020). Other several rs-fMRI studies have shown decreased FC of the TP in participants with psychiatric symptoms (Borras-Ferris et al., 2019;  Yin et al., 2018;Zhang et al., 2018b). Other studies have shown TP abnormalities, including decreased gray matter density (Zuliani et al., 2009), reduced gray matter volume (Kasai et al., 2003;Neves et al., 2015), lower mean diffusivity (reflects the microstructural integrity of white matter) (Spuhler et al., 2018), and decreased degree centrality (reflects the functional connections) (Zhou et al., 2017) in BD patients. This suggested that the abnormal structure and connectivity alterations might underlie the mood fluctuations and cognitive and emotional dysregulation that characterize BD. One task-based fMRI research on viewing affective or neutral faces tasks showed that both euthymic and manic BD patients have displayed reduced activation than HCs in the TP ( Van der Schot et al., 2010). A magnetoencephalography study investigated implicit emotional tasks that showed abnormal activation in the temporal cortices in BD (Liu et al., 2014). Thus, our findings of deficient FC between the amygdala and the TP might be related to emotional dysregulation of BD, explaining the neurobiological processes that underlie impaired emotion processing in BD.
Cytokines play a significant role in stimulating and maintaining the immune processes, and strong evidence has indicated that TNF-α could mediate the clinical presentation of BD and its neurobiological manifestations (Pantovic-Stefanovic et al., 2018). TNF-α has been described as a major inflammatory mediator, that promote the production of other cytokines such as IL-6 and C-reactive protein (Haapakoski et al., 2015). In this study, increased TNF-α levels were found in BD patients when compared to those in HCs, and this was consistent with most of the previous studies (Goldsmith et al., 2016;Niu et al., 2019). Moreover, a negative correlation was found between TNF-α levels and abnormal FC of the right medial amygdala-right MFC, and abnormal FC of the left medial amygdala-left TP in BD patients. Chronic inflammatory responses in patients with mood disorders can lead to downstream biological damage, causing damage to BD patients through upcoming actions and effects the neurotransmitter and neuropeptide systems (Sayana et al., 2017). Karabulut et al. have found that TNF-α levels were correlated with Montgomery-Asberg Depression Rating Scale scores and Clinical Global Impression Scale scores in BD patients, suggesting the progression at later stages of BD and possible association with gliosis and neuronal loss (Karabulut et al., 2019). According to an animal study, chronic unpredictable mild stress exposure induced microglia activation and overexpression of cytokine TNF-α within the MFC, and the effects were paralleled with neuronal structural changes. In contrast, chronic administration of antagonists has significantly ameliorated the dysregulation of neuronal structure and biochemical parameters within the MFC, and displayed depression-like behaviors . The cytokine (TNF-α) has been shown to cross the rodent blood-brain barrier, rapidly inducing the correlated-sensitive genes within the brain, and altering the neuronal firing rates within the amygdala and other brain regions (Davies et al., 2020). Recent literatures have found that inflammatory stimulation leads to changes in the amygdala activity and connectivity, and that the production of stress-induced inflammatory cytokines is related to the enhancement of amygdala reactivity (Mehta et al., 2018;Muscatell et al., 2015). A rs-fMRI study has found that the FC between striatum and prefrontal cortices showed negative association with serum TNF-α levels in major depressive disorder patients (Liu et al., 2019). Another fMRI study investigated that elevated levels of depression and serum cytokine levels were associated with increased regional activity in the medial frontal lobe (McIntosh et al., 2018). Evidence was obtained for distinct mechanisms by which the MFC function interacts with the inflammatory cytokine TNF-α, and this can partly explain the chronic inflammatory of impaired mood regulations in BD. This might in turn serve as a target for anti-inflammatory treatment strategies to improve emotion deficits in patients with increased inflammation, including BD McIntosh et al., 2018). However, transient elevation of inflammation might be a part of adaptation to the disease. No statistical differences of IL-6 and IL-1β were observed between the two groups. One possibility is that IL-6 and IL-1β levels were significantly increased in subjects with acute mania versus controls but there were no differences between patients with depressed BD and controls (Goldsmith et al., 2016). Yet most of our patients (76%) are in depressed mood state. Another possibility is that in this state of BD, the impacts of inflammatory cytokines might be more prominent in the regions such as the dorsolateral prefrontal cortex, insula, and lenticular nucleus, regions that might be more significantly affected during a depressed state to BD (Goldman et al., 2021). Thus, there were no association of altered FC with the clinical symptoms between the patients with BD and HCs. The possibility is that these individuals were able to maintain a depressed or euthymic status that protect them from disturbances in mood, as Lin and his colleagues previously suggested (Lin et al., 2018). For patients with BD, more research is warranted to determine if the higher TNF-α level and decreased FC of the amygdala-MFC, and -TP were associated with illness or pathogenesis to BD.
There are several limitations in this study that should be acknowledged. Firstly, patients with BD in mixed subtypes (including BD I and BD II) and mood state (including depressed and euthymic state) were examined, and it is still unclear as to whether the altered FC and TNF-α is shared by all episodes of the disease or specific to some subtypes or states of BD. Secondly, this is a cross-sectional study, lacking longitudinal data, it is not possible to determine whether these findings are trait-dependent or state-dependent framework. Longitudinal studies with large sample size will be carried out in the future. Our understanding with regard to the functional role of individual cytokines is also limited. Thirdly, the findings of TNF showed no survival after applying Bonferonni correction. Fourthly, in order to avoid excessive measurement costs, all cytokines samples were measured in one experiment, but this may not achieve high reliability and sensitivity of the retest. Fifthly, all smoking information of HCs, that affects the immune system was missed in this study. Finally, a larger sample size was needed to clarify the nature of TNF-α abnormalities in patients with BD. Also, a structural study is also warranted to confirm the results, of determining whether the changes of FC and TNF-α levels in BD patients has structural basis.

Conclusions
In conclusion, our findings of decreased FC between the right medial amygdala-right MFC, and between the left medial amygdala-left TP might reflect the disrupted corticoamygdala connectivity in patients with BD, which might show correlation with immune dysregulation in BD patients to some extent. These findings suggest the importance of dysfunctional and immune-dysregulated cortico-amygdala circuitry in the pathogenesis of BD.