Introduction
Bipolar disorder (BD) is a recurrent lifelong illness characterized by persistent fluctuations in mood states, and is associated with a poorer prognosis, greater treatment trials, higher risk of concurrent mental disorders, and higher rates of suicide (Joslyn, Hawes, Hunt, & Mitchell, Reference Joslyn, Hawes, Hunt and Mitchell2016). Depression is the most common initial presentation of BD, and for most patients a depressive episode lasts longer than a manic or hypomanic episode (Grande, Berk, Birmaher, & Vieta, Reference Grande, Berk, Birmaher and Vieta2016). It is estimated that up to 60% of bipolar depressed patients are under-recognized or misdiagnosed as recurrent unipolar depression with a mean delay of 5–10 years between illness onset and correct diagnosis (Culpepper, Reference Culpepper2014; Phillips & Kupfer, Reference Phillips and Kupfer2013). Accumulating evidence suggests that chronic inflammation and childhood adversity may be involved in the development of BD by altering brain structure (Quidé et al., Reference Quidé, Bortolasci, Spolding, Kidnapillai, Watkeys, Cohen-Woods and Green2021). Moreover, depression itself can lead to greater brain alterations in BD (Cerullo et al., Reference Cerullo, Eliassen, Smith, Fleck, Nelson, Strawn and Strakowski2014; Laidi et al., Reference Laidi, d'Albis, Wessa, Linke, Phillips, Delavest and Houenou2015; Nabulsi et al., Reference Nabulsi, McPhilemy, Kilmartin, Whittaker, Martyn, Hallahan and Cannon2020; Redlich et al., Reference Redlich, Almeida, Grotegerd, Opel, Kugel, Heindel and Dannlowski2014). Thus, identifying the brain alterations related to inflammation, childhood adversity, and clinical symptoms in bipolar II depression (BDII-D) may help to reveal the core pathophysiological mechanisms of this disorder. However, the specific relationships between these factors and brain changes in BDII-D are not yet well understood.
Increasing evidence has posited a bidirectional relationship between inflammation and mood disorders (Bauer & Teixeira, Reference Bauer and Teixeira2019). Indeed, many studies have demonstrated the association between different illness episodes of bipolar disorder and inflammatory cytokines such as Interleukin-6 (IL-6), IL-10, IL-8, and white blood cells (WBC). Meta-analyses have shown increased C-reactive protein (CRP), IL6, soluble TNF-receptor 1 (sTNF-R1), and TNF-α levels in manic and depressive episodes of BD compared to euthymic BD or healthy controls (HCs) (Bai et al., Reference Bai, Su, Tsai, Wen-Fei, Li, Pei-Chi and Mu-Hong2014; Munkholm, Braüner, Kessing, & Vinberg, Reference Munkholm, Braüner, Kessing and Vinberg2013). These findings suggest that inflammatory processes play a part in the development of BD and may even explain some of its neurobiological correlates.
Neuroimage studies have identified fronto-limbic neural circuit, fronto-striatal neural circuit, fonto-cerebellar area, insula, and temporal gyrus alterations in BD which are involved in emotion regulation, reward and cognition processing (Fournier, Chase, Almeida, & Phillips, Reference Fournier, Chase, Almeida and Phillips2016; Minichino et al., Reference Minichino, Bersani, Trabucchi, Albano, Primavera, Delle Chiaie and Biondi2014; Phillips & Swartz, Reference Phillips and Swartz2014). While inflammation has been previously implicated in BD, less is known about the role of peripheral inflammation on brain changes in BDII-D. Several studies have attempted to determine whether cytokine levels in BD differ by mood state and/or brain functional/structural alterations (Benedetti & Bollettini, Reference Benedetti and Bollettini2014; Chen et al., Reference Chen, Kao, Chang, Tu, Hsu, Huang and Bai2020a; Chen et al., Reference Chen, Chen, Chen, Zhong, Gong, Zhong and Wang2020b; Han, De Berardis, Fornaro, & Kim, Reference Han, De Berardis, Fornaro and Kim2019). A study focusing on elder BD patients found that higher peripheral levels of soluble interleukin-2 receptor (sIL-2R), sTNF-R1, and IL-1β were associated with smaller bilateral hippocampi and total gray matter volume (Tsai et al., Reference Tsai, Gildengers, Hsu, Chung, Chen and Huang2019). Another study found that a higher level of IL-10 was associated with larger left posterior cingulate volume in BD patients (Mohite et al., Reference Mohite, Salem, Cordeiro, Tannous, Mwangi, Selvaraj and Teixeira2022). Poletti et al. (Reference Poletti, Leone, Hoogenboezem, Ghiglino, Vai, de Wit and Benedetti2019) investigated the association between cytokines and brain structure and function in BD I depressed patients and found that higher levels of TNF-α and IL-8 correlated with greater cortical thickness in the anterior cingulate cortex, and markers of immune activation were associated with lower fMRI neural response to a moral valence decision task. Although the current evidence is mostly inconsistent, these studies highlight the potential involvement of inflammation damaging neural circuitry or key brain regions subserving mood and cognition.
Childhood adversity has also been associated with both inflammation levels and BD (Quidé et al., Reference Quidé, Bortolasci, Spolding, Kidnapillai, Watkeys, Cohen-Woods and Green2021). One study found that childhood adversity was associated with smaller frontal gray matter volume (GMV), but not hippocampal or amygdala volume in BD I (Begemann et al., Reference Begemann, Schutte, van Dellen, Abramovic, Boks, van Haren and Sommer2023). Another Voxel-based morphometry (VBM) study found that frontal and thalamic volumes were associated with childhood abuse in BD I patients (Duarte et al., Reference Duarte, Neves Mde, Albuquerque, de Souza-Duran, Busatto and Corrêa2016). Song et al., identified a negative correlation between precentral gyrus volume and Childhood Trauma Questionnaire (CTQ) score in BD I and II patients. Yet the role of childhood adversity on brain structural alterations in BDII-D patients remains unclear. Additionally, no studies have reported on how peripheral inflammation or childhood adversity affects brain structure in BDII-D patients.
This study aims to determine whether altered brain phenotypes in BDII-D patients are related to inflammation, childhood adversity, or psychiatric symptoms including depressed mood, anxiety, and psychosis symptoms. We hypothesize that neuroanatomical deficits will be observed in frontal-subcortical and fronto-cerebellar areas, insula, and temporal gyrus and that these regions will be related to higher inflammation, greater childhood adversity, and greater psychiatric symptoms in BDII-D.
Methods
The study was approved by the Medical Research Ethics Committee of West China Hospital, Sichuan University, and written informed consent was obtained from all participants before the study.
Participants
Three hundred and five participants (150 BDII-D from inpatient and 155 HCs) recruited for this study were between the ages of 15–65. Patients were diagnosed through an interview by two trained research psychiatrists and the diagnosis was verified using the Structured Clinical Interview for DSM-5 (American Psychiatric Association & Association, 2013). Both BDII-D patients and HCs completed the CTQ to evaluate their adverse childhood experiences. The BDII-D participants completed the 17-item Hamilton Depression Scale (HAMD), Hamilton Anxiety Scale (HAMA), and Positive and Negative Syndrome Scale (PANSS), to assess their clinical symptoms. All BDII-D patients had blood collected to test for inflammatory cytokines. Unmedicated status was given to patients who were never treated with psychotropic medication or had not used any psychotropic medication for at least the past seven days. For patients treated with psychotropic medication, we calculated medication load using a method developed by Hassel et al. (Hassel et al. Reference Hassel, Almeida, Kerr, Nau, Ladouceur, Fissell and Phillips2008). (For details on the inclusion criteria and the calculation of medication load, see the online online Supplementary Methods).
Cytokines measures
Blood was collected the day after psychiatric admission by nurses on the ward at 8:00 am and the samples were sent for testing daily to the Department of Laboratory Medicine of West China Hospital. To extract serum, the blood was allowed to clot for 30 min and then was centrifuged at 1000 g for 5 min. IL-6 was measured using an electrochemiluminescence immunoassay (Roche Diagnostics, Rotkreuz, Zug, Switzerland). IL-1β and TNF-α were measured using chemiluminescence analysis (Siemens, Erlangen, Bavaria). CRP was measured using a scatteringimmune-turbidimetric assay (Beckman Coulter, Indianapolis, IN). Blood cell samples including the absolute value of the lymphocytes, monocytes, neutrophils, platelet, and WBC count were measured using routine laboratory methods with a Hitachi 7600 (Hitachi, Tokyo, Japan). The cytokine assays were conducted in triplicate. We calculated the Neutrophil-to-Lymphocyte (NTL) ratio, Monocyte-to-Lymphocyte (MTL) ratio, and Platelet-to-Lymphocyte (PTL) ratio to reflect a low-cost, stable, reproducible, and suitable routine biomarkers of systemic inflammation (Mazza et al., Reference Mazza, Capellazzi, Tagliabue, Lucchi, Rossetti and Clerici2019; Zahorec, Reference Zahorec2001).
MRI acquisition and processing
Neuroimaging was performed on a 3.0 T magnetic resonance scanner (Siemens 3.0 T Trio Tim, Germany) with a 32-channel phased-array head coil in Huaxi MR Research Center, West China Hospital. The 3D T1-weighted images were acquired using a 3D-magnetization prepared rapid acquisition gradient echo sequences (3D-MPRAGE) with the following parameters: repetition time/echo time (TR/TE) = 2400/2.01 ms; inversion time = 1000 ms; flip angle = 8◦; slice thickness = 0.8 mm without gap; matrix = 320 × 320; field of view (FOV) = 256 × 256 mm2; voxel size = 0.8 × 0.8 × 0.8 mm3.
Voxel-based morphometry (VBM) and surface-based morphometry (SBM) analysis
The VBM and SBM analysis were processed using the Computational Anatomy Toolbox (CAT Toolbox version: 12.7, http://dbm.neuro.uni-jena.de/cat/) under Statistical Parametric Mapping (SPM12) software (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Initially, we performed the main VBM and SBM analyses between 150 BDII-D patients and 155 HCs. Additionally, we performed a subgroup analysis between 75 unmedicated BDII-D patients and HCs to verify our findings in the main analyses. (For details on how VBM and SBM analysis were performed, see the online Supplementary Methods)
Statistical analysis
Demographic and clinical data
For testing differences in demographic and clinical variables between groups, we used independent t tests for normally distributed variables, Mann–Whitney U tests for non-normally distributed variables, and χ2 tests for categorical variables. Statistical analyses were conducted using R software version 4.0.5 (https://www.rstudio.com).
VBM differences between BDII-D and HCs
Sets of general linear models (GLMs) between BDII-D and HCs using age, sex, education level, medication load, and total intracranial volume (TIV) as covariates, and a two-tailed t test was performed in SPM 12. The selection of covariates was based on a review of the literature as well as imbalanced factors between the groups. We used False Discovery Rate (FDR) correction with a p < 0.01 at the voxel level to identify the GMV differences between BDII-D patients and HCs. The extent threshold was set at 50 voxels. After VBM processing, significantly different GMVs were masked, and mean GMV values were extracted using the ROI signal extraction toolbox implemented in DPABI (Yan, Wang, Zuo, & Zang, Reference Yan, Wang, Zuo and Zang2016). VBM subgroup analyses are shown in the online Supplementary statistical analysis section.
SBM differences between BDII-D and HCs
Sets of GLMs between BDII-D and HCs using age, sex, education level, and medication load as covariates, and a two-tailed t test was performed. FDR correction with a p < 0.05 was performed at the vertex level to identify cortical differences between BDII-D and HCs. The extent threshold was set at 10 vertices. The mean values of significant differences in cortical thickness, cortical gyrification, and cortical complexity were extracted by the CAT toolbox for subsequent analysis. SBM subgroup analyses are shown in the online Supplementary statistical analysis section.
Partial Pearson correlation analysis
Partial Pearson correlations were performed for variables with a normal distribution to explore the association between childhood adversity and brain phenotypes. Associations between inflammation, psychiatric symptoms, and brain structure were performed in the BDII-D group alone. Age, sex, education level, medication load, and TIV were used as covariates for VBM measures, while age, sex, education level, and medication load were used for SBM measures. The ‘stats’ R package was used to perform FDR-correction and q values were reported (Benjamini & Hochberg, Reference Benjamini and Hochberg1995). To define the number of independent tests we report on correlations using each inflammatory cytokine with 15 GMVs or eight cortical thicknesses resulting in 15 and eight pairs of comparisons. A similar procedure was used between clinical scores and either 15 GMVs or eight cortical thicknesses.
Results
Cohort characteristics
BDII-D patients demonstrated significant statistical differences in age (22.75 ± 7.24 v. 25.51 ± 7.15, Cohen's d = −0.38, p = 0.001), education levels (13.79 ± 2.31 v. 16.90 ± 3.15, Cohen's d = −1.13, p < 0.001), and CTQ score (47.01 ± 17.21 v. 28.41 ± 5.49, Cohen's d = 1.46, p < 0.001) compared with HCs. No differences were found between BDII-D patients and HCs for sex, marital status or childbearing status. Seventy-five BDII-D patients were categorized as unmedicated. Fifty-four patients received antidepressant treatment; 12 patients received serotonin and norepinephrine reuptake inhibitors (SNRIs), 22 received selective serotonin reuptake inhibitors (SSRIs), and 20 received agomelatine. Sixty patients received mood stabilizers, 50 patients received antipsychotics, and 27 patients received benzodiazepines. Further information is shown in Table 1.
Table 1. Demographic and clinical characteristics of participants
Abbreviations: BD, bipolar disorder; HAMD, 17-item Hamilton Depression Scale; HAMA, Hamilton Anxiety Scale; PANSS, Positive and Negative Syndrome Scale; P, positive; N, negative; G, general; T, Total; CTQ, Childhood Trauma Questionnaire; IL, Interleukin; CRP, C-reactive protein; TNF, Tumor Necrosis Factor; WBC, white blood cell; SNRIs, serotonin and norepinephrine reuptake inhibitors; SSRIs, selective serotonin reuptake inhibitors.
Group differences in brain phenotypes
BDII-D patients compared with HCs demonstrated decreased GMVs in the right inferior cerebellum, right superior frontal gyrus (medial SFG), right middle frontal gyrus (orbital MFG), right inferior frontal gyrus (triangular IFG), right insula, left middle temporal gyrus (MTG), left opercular IFG, left rectus, bilateral temporal pole, bilateral MFG, bilateral orbital IFG, and bilateral caudate (Fig. 1a and online Supplementary Table S1). Compared with HCs, BDII-D patients showed significantly reduced cortical thickness in the left inferior temporal, left lateral orbitofrontal, right postcentral, right superior frontal, bilateral caudal middle frontal, and bilateral rostral middle frontal areas (Fig. 1b and online Supplementary Table S2). No significant differences were identified in cortical gyrification or complexity between BDII-D patients and HCs. Subgroup analyses between 75 unmedicated BDII-D patients and HCs are shown in online Supplementary Result S1.
Figure 1. (a) Voxel-based morphometry results of gray matter volume in BDII-D (n = 150) patients compared with healthy controls (n = 155) using age, sex, education level, medication load, and total intracranial volume as covariates (Slices are presented from left to right). Voxel level p value < 0.01 (FDR corrected), voxels size >50. The color bar presented T value. (b) Surface-based morphometry results of cortical thickness in BDII-D (n = 150) patients compared with healthy controls (n = 155) using age, sex, education level, and medication load as covariates. Vertex level p value < 0.05 (FDR corrected), vertices size > 10. The color bar presented p value.
Brain phenotype correlations with childhood adversity in BDII-D and HCs
CTQ was used to evaluate emotional abuse, physical abuse, sexual abuse, emotional neglect, physical neglect, and total childhood adversity experienced in BDII-D and HCs participants before 16 years of age. Significant correlations between childhood adversity and neuroanatomical alterations were found in BDII-D patients but not in HCs. Specifically, significant correlations were found between higher sexual abuse scores and smaller right MFG volume (r = −0.348, q < 0.001) (Fig. 2a). Significant correlations between larger right orbital MFG volume and greater physical neglect score (r = 0.254, q = 0.03) (Fig. 2a) were identified.
Figure 2. Partial correlations between childhood adversity and significantly altered gray matter volumes (a) and cortical thicknesses (b) in BDII-D and HCs group. Summary of brain regions impacted by childhood adversity (c). Partial correlations were pre-adjusted for age, sex, education level, and medication load. Volume data were additionally adjusted for total intracranial volume. Uncorrected p values are presented, and only relationships between greater sexual abuse and smaller GMV of right MFG (q < 0.001), as well as greater physical neglect and larger GMV of right orbital MFG (q = 0.03), survived FDR correction (highlighted with red stars). Abbreviations: BDII, D-bipolar II depression; HCs, healthy controls; R, right; L, left; CTQ, Childhood Trauma Questionnaire.
The preliminary findings demonstrating the relationships between childhood trauma and brain volume or cortical thickness are shown in online Supplementary Result S2 and Figs 2a and b. Figure 2c summarizes brain regions that may be affected by childhood adversity in our study.
Brain phenotype correlations with inflammatory cytokines in BDII-D
There were several negative relationships between inflammatory cytokines and neuroanatomical alterations in BDII-D. There were significant associations between higher IL-6 (r = −0.266, q = 0.015) levels and smaller GMV in the left MFG (Fig. 3a). Higher counts of WBC were found to be significantly associated with smaller GMVs in the right inferior cerebellum (r = −0.227, q = 0.015) after multiple comparison corrections (Fig. 3b). There were also several associations between inflammatory cytokines and brain phenotypes, but they did not survive multiple comparison corrections (online Supplementary Results S3 and Supplementary Figs S1 and S2). The correlational data can be found in the online Supplementary Tables S3 and S4.
Figure 3. Partial correlations between IL-6, WBC count, and gray matter volume in bipolar II depression. Partial correlations were adjusted for age, sex, education level, medication load, and total intracranial volume. Correlations between higher counts of WBC and smaller GMV in the right inferior cerebellum and higher IL-6 and smaller GMV in the left MFG survived FDR correction.
Brain phenotype correlations with psychiatric symptoms in BDII-D
There were positive relationships between PANSS and GMV in several brain regions. A higher PANSS positive score was significantly correlated with larger GMVs of the left middle temporal pole (r = 0.245, p = 0.002, q = 0.03; Fig. 4). Associations between GMV and psychiatric symptoms (including depressed mood, anxiety mood, and psychotic symptoms) in BD II-D are shown in online Supplementary Results S4 and Supplementary Fig. S3.
Figure 4. Partial correlations between PANSS positive total symptom score and gray matter volume of the left middle temporal pole in bipolar II depression. Partial correlations were adjusted for age, sex, education level, medication load, and total intracranial volume. The correlation survived FDR correction.
There were no significant associations between GMVs and the duration of depressive episodes, duration of illness, age at onset, number of hypomanic episodes, PANSS general score, physical abuse score, TNF-α, MTL ratio, and PTL ratio. No association was found between cortical thickness and the duration of depressive episodes, duration of illness, age at onset, number of hypomanic episodes, psychiatric symptoms (HAMD score, HAMA score, PANSS score), physical abuse score, emotional abuse score, IL-1β, TNF-α, NTL ratio, MTL ratio, and PTL ratio.
Discussion
We found that compared to HCs, BDII-D patients had smaller GMV which was primarily found in frontostriatal and fronto-cerebellar areas, insula, rectus, and temporal gyrus, and lower cortical thickness in frontal and temporal areas. We demonstrated that inflammatory cytokines (including IL-1β, IL-6, CRP, WBC counts, and NTL ratio), childhood adversity, and psychiatric symptoms were associated with neuroanatomical alterations in BDII-D, although only a few results survived multiple comparison corrections. Specifically, significant correlations between higher WBC and smaller right inferior cerebellum volume (r = −0.227, q = 0.015) as well as higher IL-6 and smaller left MFG volume (r = −0.266, q = 0.015). We also uncovered a significant relationship between greater sexual abuse and smaller GMV of the right MFG, as well as greater physical neglect and larger GMV of the right orbital MFG. Additionally, we found that more severe positive symptom was significantly correlated with larger left middle temporal volume. These results support our hypothesis that neuroanatomical alterations in frontal-subcortical and fronto-cerebellar areas, insula, and temporal gyrus in BDII-D patients may be related to inflammation, trauma, and psychosis symptoms.
Studies indicated that microglia and neurons maintain bidirectional communication through the interaction of CX3CLR1 and CX3CL1, regulating their respective functions during stable states (Wang et al., Reference Wang, He, Sun, Ren, Liu, Wang and Yang2022). If this process becomes disrupted it can lead to altered microglia, synaptic pruning, and neural circuitry. Inflammation has been implicated in brain structural alterations underlying neuropsychiatric disorders via microglia mediated synaptic pruning, neurogenic dysfunction leading to GMV changes (Irwin & Cole, Reference Irwin and Cole2011; Valkanova, Ebmeier, & Allan, Reference Valkanova, Ebmeier and Allan2013; Williams et al., Reference Williams, Burgess, Suckling, Lalousis, Batool, Griffiths and Upthegrove2022). In general, inflammation is a complex biological response that involves the activation of various immune cells, the release of cytokines and other inflammatory molecules, and the recruitment of immune cells to sites of injury or infection. Inflammatory markers can provide information about the various aspects of the inflammatory response and thus may be more or less relevant to understanding the relationship between inflammation and brain structure. Previous studies have collectively suggested that elevated immune-inflammatory signaling is a mechanism that is relevant to the patho-etiology of BD (Modabbernia, Taslimi, Brietzke, & Ashrafi, Reference Modabbernia, Taslimi, Brietzke and Ashrafi2013), however, the associations between inflammation and brain phenotypes of BDII-D are still unknown. Studies and meta-analyses have indicated that the IL-6, TNF-α, and CRP were elevated in major depressive disorder and BD, while IL-1β varied by the specific mood episode (Miller, Haroon, Raison, & Felger, Reference Miller, Haroon, Raison and Felger2013; Solmi et al., Reference Solmi, Suresh Sharma, Osimo, Fornaro, Bortolato, Croatto and Carvalho2021). Of these, IL-6 exhibits the most consistent trait function, showing stable differences throughout various episodes, including euthymia (Solmi et al., Reference Solmi, Suresh Sharma, Osimo, Fornaro, Bortolato, Croatto and Carvalho2021). IL-6 exhibits pleiotropic activity within the bloodstream, triggering a range of effects, such as stimulating the production of acute-phase proteins (e.g. CRP), activating the acquired immune response by promoting antibody production and the development of effector T-cells, and by encouraging the differentiation or proliferation of non-immune cells (Goldsmith, Rapaport, & Miller, Reference Goldsmith, Rapaport and Miller2016). One prospective study investigated whether inflammatory cytokines change with vacillating mood states in bipolar I disorder over 6–12 months indicating that the hs-CRP decreased by 40% during depressive states compared to euthymia, and decreased by 48% compared to hypomanic/manic states, while other cytokines including BDNF, IL-6, IL-8, IL-18, and TNF-α were not found to vary in accordance with affective states (Jacoby, Munkholm, Vinberg, Pedersen, & Kessing, Reference Jacoby, Munkholm, Vinberg, Pedersen and Kessing2016). IL-6 and CRP identified herein are important since they are involved in stress-related immune and vascular dysfunction that can negatively impact the central nervous system (Goldsmith et al., Reference Goldsmith, Rapaport and Miller2016). Although our findings preliminarily identified associations between increased CRP, IL-1β, IL-6, WBC counts, NTL ratio and decreased GMV in caudate, insula, cerebellum, temporal pole, and frontal areas in BDII-D patients, as well as relationships between higher WBC, CRP, and IL-6 and lower cortical thickness in frontal areas, it was only higher WBC counts and smaller right inferior cerebellum volume and higher IL-6 and left MFG volume that survived multiple comparisons correction. A Mendelian randomization study of neuropsychiatric disorders predicted the level of IL-6 but not CRP was associated with GMV in MTG and SFG (Williams et al., Reference Williams, Burgess, Suckling, Lalousis, Batool, Griffiths and Upthegrove2022). Tsai et al. (Reference Tsai, Gildengers, Hsu, Chung, Chen and Huang2019) found that the right hippocampal volume was negatively associated with sIL-2R and sTNF-R1 levels, while the left hippocampal volume was negatively associated with sIL-2R levels in elderly euthymic bipolar I disorder patients. They indicated that neuroinflammation and the pathophysiology of BD may play a role in the neuroprogression of BD. Despite the different brain regions associated with inflammation, this inverse relationship between inflammatory cytokines and brain phenotypes in the fronto-cerebellar area are consistent with previous studies and supports our hypothesis.
Considering the high prevalence of childhood adversity in mood disorders (Palmier-Claus, Berry, Bucci, Mansell, & Varese, Reference Palmier-Claus, Berry, Bucci, Mansell and Varese2016), studying the links between childhood adversity experienced with brain structural alterations in BDII-D could have significant clinical implications. The mechanism underlying the processes of how childhood trauma affects brain structure in bipolar depression appears to be complex. The prevailing theory is that stress-induced neurodevelopmental changes can lead to dysregulation of catecholamines and the hypothalamic–pituitary–adrenal (HPA) axis. Stress also stimulates chronic low-grade inflammation, which in turn promotes the breakdown of tryptophan to kynurenine, kynurenic acid and quinolinic acid, resulting in activation of the kynurenine-tryptophan metabolic pathway, leading to a decreased serotonin and increased glutamatergic neurotransmitters in the brain, which is believed to produce brain changes (Heim, Newport, Mletzko, Miller, & Nemeroff, Reference Heim, Newport, Mletzko, Miller and Nemeroff2008; McKay et al., Reference McKay, Cannon, Chambers, Conroy, Coughlan, Dodd and Clarke2021; Rosenblat et al., Reference Rosenblat, Brietzke, Mansur, Maruschak, Lee and McIntyre2015). Studies observed that BD patients who were exposed to high childhood adversity had GMV deficits in the orbitofrontal gyrus, insula, and thalamus (Duarte et al., Reference Duarte, Neves Mde, Albuquerque, de Souza-Duran, Busatto and Corrêa2016; Poletti et al., Reference Poletti, Vai, Smeraldi, Cavallaro, Colombo and Benedetti2016). Similarly, the associations between childhood adversity and brain alterations in the frontal area and frontal-limbic network in BD patients were also reported in several studies (Begemann et al., Reference Begemann, Schutte, van Dellen, Abramovic, Boks, van Haren and Sommer2023; Hsieh et al., Reference Hsieh, Wu, Tseng, Wei, Huang, Chang and Chen2021; Souza-Queiroz et al., Reference Souza-Queiroz, Boisgontier, Etain, Poupon, Duclap, d'Albis and Houenou2016). Besides, a task-based MRI study found that childhood trauma is associated with increased brain responses to emotionally negative stimuli in several brain regions including the temporal gyrus. However, few studies have been able to study these interrelationships in BDII-D populations. Our study found a tendency for childhood adversity to alter GMV in the right frontal areas and insula, as well as the cortical thickness in the left inferior temporal cortex in BDII-D. Interestingly, our finding observed the right orbital MFG and MFG volumes deficits were significantly associated with sexual abuse and physical neglect, respectively. In line with previous human or animal findings, our study also provided evidence that GMV in frontal areas may be particularly sensitive to childhood trauma (Begemann et al., Reference Begemann, Schutte, van Dellen, Abramovic, Boks, van Haren and Sommer2023; Schubert, Porkess, Dashdorj, Fone, & Auer, Reference Schubert, Porkess, Dashdorj, Fone and Auer2009). Although complex brain functioning is related to cognition and emotional behavior, the frontal area is generally implicated in emotion regulation and the top-down generation of emotions with increasing cognitive complexity (Ochsner et al., Reference Ochsner, Ray, Hughes, McRae, Cooper, Weber and Gross2009). Childhood adversity-related cognitive and affective deficits may be mediated by abnormal brain structural changes, mainly in the frontal cortex. However, addressing the mechanism behind these correlations warrants further examination. Besides, it is not known yet if these right-lateralized impairments of GMVs and left-lateralized impairments of cortical thickness found in the current study imply an accidental or particular manifestation in BDII-D.
The relationships between psychiatric symptoms and brain phenotypes are discussed in the online Supplementary Discussion.
Limitations
This study had several limitations worth noting. First, we did not collect inflammatory cytokines in HCs, which makes it difficult to determine if the inverse correlation between the inflammatory measurements and structural brain phenotypes are specific to the BD-II group. Additionally, our study included inflammatory cytokines that were measured during an acute episode of depression and thus don't reflect long-term inflammatory changes in BDII-D. Our subsequent study will address this issue by following patients through different episodes and recording and comparing the levels of inflammatory cytokines during different episodes. Second, the CTQ we used in our study may not precisely reflect the individual's adverse childhood experiences because of its retrospective nature, although it has been widely used in research for clinical and non-clinical individuals. Third, some correlations did not survive multiple comparison correction. For that reason, we consider some correlations in this analysis as preliminary. Fourth, because of the cross-sectional nature of our study, we cannot accurately explain how childhood trauma exerts its ongoing influence on the adult brain, nor comment on the causal relationship between inflammation, childhood trauma, clinical symptoms, and brain structural changes in BDII-D. Last, our study included part of adolescents, however, we normalized T1 images into MNI space but not into individual template created by our own data. We will take this factor into careful consideration for future studies.
Conclusion
Our study identified GMV differences in frontostriatal and fronto-cerebellar area, insula, rectus, and temporal gyrus as well as cortical thickness differences in frontal and temporal areas between BDII-D patients and HCs. Our study explored the potential associations between brain morphometrical phenotypes and inflammation, childhood adversity, and psychiatric symptoms in BDII-D populations. Further studies could explore the precise neurobiological mechanisms involved across various stages of the illness.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723002477
Acknowledgments
We thank all participants for their kind contributions. Figure 2c was drawn partly by Figdraw. Yuan Cao thanks the China Scholarship Council.
Authors’ contributions
Y. C. wrote the first draft of the paper; P. L. and H. S. edited the paper; Y. C., P. L., C. J. Q., and Z. Y. J. designed the research; C. J. Q., Q. Y. G, and Z. Y. J. supervised the study conduction. Y. C., H. S., G. J. D., S. Y. L., and H. Q. X. were involved in clinical data collection. Y. C., H. S. X, J. S. M., X. P. L., and B. L. W. participated in MRI scans; Y. C., H. S., and X. Q. Z analyzed data.
Funding statement
This study was supported by the National Natural Science Foundation of China (Grant Nos. 82271947 and 81971595 to Dr Jia), the Key Program of Natural Science Foundation of Sichuan Province (Grant No. 2022NSFSC0047 to Dr Jia), the Key R&D Support Plan of Chengdu Science and Technology Bureau (Grant No. 2022-YF05-01766-SN to Dr Jia), the 1.3.5 Project for Disciplines of Excellence–Clinical Research Incubation Project, West China Hospital, Sichuan University (Grant No. 2020HXFH005 to Dr Jia, Grant No. ZYJC21083 to Dr Qiu, and Grant No. 2022HXFH029 to Dr Qiu), and the Department of Science and Technology of Sichuan provincial government (Grant No. 2022YFS0345 to Dr Qiu).
Competing interests
None.
